{"api":"v1","id":"chips.inference","release":{"id":"2026-11-10","date":"2026-11-10","checked":"2026-11-10","sha256":"bb6f682d109ff6e73907d02eb172bf77eff62374b1c41d0376279448738393cf"},"csv":"https://jesseahawkins.com/data/releases/chips/inference/2026-11-10.csv","fields":[{"name":"round","type":"number"},{"name":"result_id","type":"string"},{"name":"submitter","type":"string"},{"name":"system","type":"string"},{"name":"accelerator","type":"string"},{"name":"chip","type":"string"},{"name":"chips","type":"number"},{"name":"chips_note","type":"string"},{"name":"benchmark","type":"string"},{"name":"benchmark_name","type":"string"},{"name":"scenario","type":"string"},{"name":"result","type":"number"},{"name":"units","type":"string"},{"name":"status","type":"string"},{"name":"per_chip","type":"number"},{"name":"per_chip_status","type":"string"},{"name":"published","type":"string"},{"name":"source_url","type":"string"},{"name":"kept_copy","type":"string"},{"name":"kept_sha256","type":"string"},{"name":"page","type":"string"},{"name":"license","type":"string"},{"name":"trademark","type":"string"}],"rowCount":1441,"rows":[{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":8840.42,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1105.0525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27803.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3475.4875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":24593.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3074.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":8840.42,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1105.0525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27803.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3475.4875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24593.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3074.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":53017.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6627.1625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0001","submitter":"AMD","system":"Supermicro AS-8125GS-TNMR2","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":47804.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5975.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34520.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4315.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":32027.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4003.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34520.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4315.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":32027.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4003.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":68781.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8597.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0002","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":59783.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7472.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0003","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":18407.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2300.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0003","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34555,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4319.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0003","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":18407.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2300.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0003","submitter":"AMD","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34555,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4319.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0004","submitter":"AMD","system":"Quanta S7PA","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":18.5856,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.3232,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0004","submitter":"AMD","system":"Quanta S7PA","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":16.2044,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.02555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":580460,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":72557.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":450100,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":56262.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":348225,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":43528.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":330555,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":41319.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":30487,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3810.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":25614.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3201.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":30490.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3811.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24554.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3069.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":13319.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1664.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0005","submitter":"ASUSTeK","system":"ESC8000A-E12 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":12801.9,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1600.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":15566.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1945.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":33987.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4248.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31241,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3905.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":15566.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1945.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":33987.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4248.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31241,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3905.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":68121.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8515.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0006","submitter":"ASUSTeK","system":"ESC_A8A_MI325X_256GBx8","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":55800.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6975.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":643122,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":80390.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":591162,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":73895.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":388058,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":48507.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":370066,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":46258.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":35075.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4384.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":34193.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4274.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34887.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4360.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":29026.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3628.3625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":61372.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7671.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":59992.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7499.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14451.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1806.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0007","submitter":"ASUSTeK","system":"ASUSTeK ESC N8 H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":14406,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1800.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0008","submitter":"Azure","system":"ND_GB200_v6","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":52061.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13015.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0008","submitter":"Azure","system":"ND_GB200_v6","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":46016.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11504.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0008","submitter":"Azure","system":"ND_GB200_v6","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":847.233,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":211.80825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0008","submitter":"Azure","system":"ND_GB200_v6","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":584.008,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":146.002,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0009","submitter":"Broadcom_Supermicro","system":"Supermicro SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1650.43,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":206.30375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0009","submitter":"Broadcom_Supermicro","system":"Supermicro SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1243.86,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.4825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0009","submitter":"Broadcom_Supermicro","system":"Supermicro SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":31.8413,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.9801625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0009","submitter":"Broadcom_Supermicro","system":"Supermicro SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.7042,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.588025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0009","submitter":"Broadcom_Supermicro","system":"Supermicro SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":41927.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5240.925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0010","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":498.421,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0010","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":400.421,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0010","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1403.84,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0011","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":50776.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6347.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0011","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":38696.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4837.1125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0011","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":13263.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1657.925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0011","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":13202.8,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1650.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x L40S-48GB, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":6518.39,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":814.79875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x L40S-48GB, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":6101.41,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":762.67625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":68894.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4305.925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":66439,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4152.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":68894.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4305.925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":66439,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4152.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1086.01,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":67.875625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0013","submitter":"Cisco","system":"Cisco UCS C885A M8 (16x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":16,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":584.291,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":36.5181875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":19621.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2452.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34548.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4318.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33163.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4145.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":19621.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2452.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34548.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4318.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33163.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4145.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":58304.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7288.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":41291,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5161.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14923.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1865.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":14355.8,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1794.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":19.1227,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.3903375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0014","submitter":"Cisco","system":"Cisco UCS C885A M8 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":18.3718,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.296475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":62351.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3896.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":61101.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3818.83125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":62351.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3896.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":61101.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3818.83125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":783.131,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":48.9456875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0015","submitter":"Cisco","system":"HPF HGX System (16xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":16,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":563.365,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":35.2103125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":124879,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3902.46875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":122274,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3821.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":124879,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3902.46875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":122274,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3821.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1561.57,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":48.7990625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0016","submitter":"Cisco","system":"HPF HGX System (32xH100-SXM-80GB, TensorRT)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":32,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1126.94,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":35.216875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":12503.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11801.7,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":819.624,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":257.749,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":501.263,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":400.419,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":16102.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0017","submitter":"Cisco","system":"X210M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1418.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0018","submitter":"Dell","system":"PowerEdge R470 (2x L4-PCIE-24GB, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":2,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":458.511,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":229.2555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0018","submitter":"Dell","system":"PowerEdge R470 (2x L4-PCIE-24GB, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":2,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":428.115,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":214.0575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0019","submitter":"Dell","system":"PowerEdge R570 (4x L4-PCIE-24GBx4, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":4,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":905.51,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":226.3775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0019","submitter":"Dell","system":"PowerEdge R570 (4x L4-PCIE-24GBx4, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":4,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":813.985,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":203.49625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0019","submitter":"Dell","system":"PowerEdge R570 (4x L4-PCIE-24GBx4, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":4,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":1.00634,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.251585,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0019","submitter":"Dell","system":"PowerEdge R570 (4x L4-PCIE-24GBx4, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":4,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":0.880856,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":0.220214,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0019","submitter":"Dell","system":"PowerEdge R570 (4x L4-PCIE-24GBx4, TensorRT)","accelerator":"NVIDIA L4-PCIe-24GB","chip":"NVIDIA L4","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1397.01,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":349.2525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0020","submitter":"Dell","system":"R770_XEON6787Px2","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":12421.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0020","submitter":"Dell","system":"R770_XEON6787Px2","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11801.8,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0021","submitter":"Dell","system":"Dell PowerEdge R7715 (3x H100-NVL-94GB, TensorRT)","accelerator":"NVIDIA H100-NVL-94GB","chip":"NVIDIA H100 NVL","chips":3,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":8301.9,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2767.2999999999997,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0022","submitter":"Dell","system":"Dell PowerEdge XE7740 (8x L40S, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":6509.66,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":813.7075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0022","submitter":"Dell","system":"Dell PowerEdge XE7740 (8x L40S, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":6101.31,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":762.66375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0022","submitter":"Dell","system":"Dell PowerEdge XE7740 (8x L40S, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":5.52041,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.69005125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0022","submitter":"Dell","system":"Dell PowerEdge XE7740 (8x L40S, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":5.9027,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":0.7378375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0022","submitter":"Dell","system":"Dell PowerEdge XE7740 (8x L40S, TensorRT)","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":9272,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1159,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":16419.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2052.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":31267.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3908.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":29070.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3633.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":16419.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2052.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":31267.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3908.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":29070.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3633.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":13421.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1677.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":12601.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1575.1625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":17.0907,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.1363375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":16.2171,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.0271375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0023","submitter":"Dell","system":"Dell PowerEdge XE7745 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":31674.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3959.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":30493.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3811.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":13913.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1739.1625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":60137.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7517.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":101253,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12656.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99139.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12392.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":60137.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7517.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101253,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12656.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99139.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12392.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":751.369,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.921125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1648.03,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":206.00375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1243.17,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.39625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":102907,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12863.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":143309,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17913.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128787,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16098.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0024","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":45024.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5628.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34588.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4323.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33161.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4145.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34588.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4323.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33161.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4145.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14542.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1817.85,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":13603.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1700.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0025","submitter":"Dell","system":"Dell PowerEdge XE9680L (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":34014.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4251.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":21915.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2739.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":35316.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4414.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33244.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4155.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":21915.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2739.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":35316.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4414.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33244.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4155.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14960.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1870.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0026","submitter":"Dell","system":"Dell PowerEdge XE9680 (8xH200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":13603.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1700.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":8455.14,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1056.8925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27325.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3415.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":24747.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3093.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":8455.14,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1056.8925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27325.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3415.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24747.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3093.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":53333.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6666.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0027","submitter":"Dell","system":"Dell Poweredge XE9680","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":47801.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5975.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":30766.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3845.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":13963,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1745.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":102131,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12766.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99214.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12401.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":102131,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12766.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99214.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12401.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":751.589,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.948625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1660.71,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":207.58875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1246.56,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.82,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":143904,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17988,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128794,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16099.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":32.2916,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4.03645,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4651,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5581375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0028","submitter":"Dell","system":"Dell PowerEdge XE9685L (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":44922.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5615.3,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34301.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4287.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33384.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4173.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34485.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4310.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33370.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4171.325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":547.261,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":68.407625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":277.331,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":34.666375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":18.6415,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.3301875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0029","submitter":"Dell_Broadcom","system":"Dell PowerEdge XE9680 (8x H200-SXM-141GB,VMware ESXi 9.0.0)","accelerator":"Virtualized NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":17.9503,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.2437875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0030","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (2x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":53572.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3348.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0030","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (2x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":47561.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2972.60625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0030","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (2x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":53572.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3348.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0030","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (2x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":47561.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2972.60625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0031","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27678.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3459.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0031","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":24532.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3066.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0031","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27678.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3459.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0031","submitter":"Dell_MangoBoost","system":"Dell PowerEdge XE9680 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24532.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3066.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0032","submitter":"GATEOverflow","system":"GATE Overflow Intel Sapphire Rapids (2x RTX 4090)","accelerator":"NVIDIA GeForce RTX 4090","chip":"NVIDIA GeForce RTX 4090","chips":2,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":1760.45,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":880.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0032","submitter":"GATEOverflow","system":"GATE Overflow Intel Sapphire Rapids (2x RTX 4090)","accelerator":"NVIDIA GeForce RTX 4090","chip":"NVIDIA GeForce RTX 4090","chips":2,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":1448.97,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":724.485,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0033","submitter":"GATEOverflow","system":"GATE Overflow Intel Sapphire Rapids (2x RTX 4090)","accelerator":"NVIDIA GeForce RTX 4090","chip":"NVIDIA GeForce RTX 4090","chips":2,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":104.953,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":52.4765,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":18825.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2353.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34483.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4310.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":32139.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4017.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":18825.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2353.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34483.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4310.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":32139.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4017.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":69692,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8711.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0045","submitter":"GigaComputing","system":"G893-ZX1-AAX2","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":58242.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7280.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":30873.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3859.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":15977.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1997.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":62851.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7856.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":98607.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12325.9625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99066,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12383.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":62851.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7856.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":98607.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12325.9625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99066,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12383.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":719.409,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":89.926125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1625.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":203.2075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1244.97,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.62125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":102829,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12853.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":142136,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17767,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128660,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16082.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":632483,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":79060.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":31.5784,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.9473,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4612,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.55765,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0046","submitter":"GigaComputing","system":"G894-SD1","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":42565.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5320.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":31920.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3990.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":14630.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1828.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":52140.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6517.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":78463.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9807.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99169.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12396.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":52140.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6517.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":78463.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9807.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99169.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12396.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":750.447,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.805875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1655.36,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":206.92,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1246.79,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.84875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":141138,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17642.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":119628,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14953.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":630364,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":78795.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":32.1157,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4.0144625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4633,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5579125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0047","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":42755.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5344.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":19554.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2444.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":35043.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4380.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":19554.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2444.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":35043.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4380.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14718.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1839.85,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":13603.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1700.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0048","submitter":"Google","system":"NVIDIA H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":34088.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4261.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":20432.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2554.0125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34964.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4370.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33164.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4145.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":20356.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2544.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34797,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4349.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33144.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4143.0125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":53475.2,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":6684.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":66036.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8254.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":64914.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8114.325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":62108.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7763.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":60955.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7619.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14996.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1874.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":14205.6,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1775.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0049","submitter":"HPE","system":"HPE Cray XD670 (8x H200-SXM-141GB)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":34450.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4306.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x H200-NVL-141GB)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":10,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":713817,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":71381.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x H200-NVL-141GB)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":10,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":650211,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":65021.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x H200-NVL-141GB)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":10,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":433591,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":43359.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x H200-NVL-141GB)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":10,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":413565,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":41356.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":26205.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3275.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":26005.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3250.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":26205.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3275.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":26001,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3250.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":36052.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":4506.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":46841.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5855.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":46060,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5757.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":32738.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4092.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":30134.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3766.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":11.0758,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1.384475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0051","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":10.8825,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.3603125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x H200-NVL-141GB)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":2,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":7925.55,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3962.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0053","submitter":"HPE","system":"HPE ProLiant Compute DL384 Gen12 (2x GH200-144GB_aarch64)","accelerator":"NVIDIA GH200 Grace Hopper Superchip 144GB","chip":"NVIDIA GH200","chips":2,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":174456,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":87228,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0053","submitter":"HPE","system":"HPE ProLiant Compute DL384 Gen12 (2x GH200-144GB_aarch64)","accelerator":"NVIDIA GH200 Grace Hopper Superchip 144GB","chip":"NVIDIA GH200","chips":2,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":161030,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":80515,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":18495.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":18190.1,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":18495.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":18190.1,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1232.89,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":450.836,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":708.252,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":579.508,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":24121.9,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0055","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2187.39,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":3009.38,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":752.345,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":2110.05,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":527.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":3009.38,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":752.345,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":2110.05,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":527.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":6472.37,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1618.0925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0056","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"[MAXSUN] MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":5348.45,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1337.1125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":12274.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11801.6,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":797.84,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":289.973,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":479.573,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":365.415,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":13847.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0058","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_BareMetal","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1481.52,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":11981.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11566.1,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":774.122,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":260.99,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":467.451,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":361.651,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":12771,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0059","submitter":"Intel_VMware","system":"1-node-2S-GNR_86C_ESXi_172VCPU-VM","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1420.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":102725,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12840.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99993.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12499.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1648.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":206.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1246.79,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.84875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":32.5691,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4.0711375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0060","submitter":"Lambda","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4629,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5578625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0061","submitter":"Lenovo","system":"ThinkSystem SR680a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":102909,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12863.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0061","submitter":"Lenovo","system":"ThinkSystem SR680a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99159.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12394.9125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0061","submitter":"Lenovo","system":"ThinkSystem SR680a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":102909,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12863.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0061","submitter":"Lenovo","system":"ThinkSystem SR680a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99203.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12400.4875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0061","submitter":"Lenovo","system":"ThinkSystem SR680a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":43698.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5462.325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":752.624,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":94.078,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1637.11,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":204.63875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1249.04,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":156.13,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":146960,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18370,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128689,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16086.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":618486,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":77310.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":31.7836,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.97295,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4637,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5579625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0062","submitter":"Lenovo","system":"ThinkSystem SR780a V3 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":44365.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5545.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":12230,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11991.5,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":776.849,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":275.781,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":468.732,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":375.464,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":13621.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0063","submitter":"Lenovo","system":"ThinkSystem_SR650_V4_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1411.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0065","submitter":"MangoBoost","system":"MangoBoost Heterogeneous Cluster (4x8x MI300X-192GB, 2x8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x32), AMD Instinct MI325X 256GB HBM3e (x16)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":169197,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0065","submitter":"MangoBoost","system":"MangoBoost Heterogeneous Cluster (4x8x MI300X-192GB, 2x8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x32), AMD Instinct MI325X 256GB HBM3e (x16)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":153076,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0065","submitter":"MangoBoost","system":"MangoBoost Heterogeneous Cluster (4x8x MI300X-192GB, 2x8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x32), AMD Instinct MI325X 256GB HBM3e (x16)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":169197,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0065","submitter":"MangoBoost","system":"MangoBoost Heterogeneous Cluster (4x8x MI300X-192GB, 2x8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x32), AMD Instinct MI325X 256GB HBM3e (x16)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":153076,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0066","submitter":"MangoBoost","system":"Supermicro AS-8125GS-TNMR2 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27854.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3481.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0066","submitter":"MangoBoost","system":"Supermicro AS-8125GS-TNMR2 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":24860.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3107.5625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0066","submitter":"MangoBoost","system":"Supermicro AS-8125GS-TNMR2 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27854.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3481.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0066","submitter":"MangoBoost","system":"Supermicro AS-8125GS-TNMR2 (8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24860.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3107.5625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0067","submitter":"MangoBoost","system":"MangoBoost (8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34454.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4306.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0067","submitter":"MangoBoost","system":"MangoBoost (8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31671.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3958.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0067","submitter":"MangoBoost","system":"MangoBoost (8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34454.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4306.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0067","submitter":"MangoBoost","system":"MangoBoost (8x MI325X-256GB, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31671.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3958.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":18846.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2355.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":33703,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4212.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31755.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3969.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":18846.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2355.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":33703,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4212.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31755.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3969.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":67964.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8495.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":57087.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7135.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":33379,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4172.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":18592.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2324.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":59545.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7443.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":101527,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12690.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99123,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12390.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":59545.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7443.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101527,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12690.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99123,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12390.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":751.366,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.92075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1624.11,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":203.01375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1240.42,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.0525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":102903,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12862.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":136752,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17094,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128541,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16067.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":651230,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":81403.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":31.7754,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.971925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4631,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5578875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0069","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":45333.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5666.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":29745.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7436.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":51736.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12934.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":49359.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12339.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":29745.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7436.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":51736.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12934.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0070","submitter":"NVIDIA","system":"(4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":49359.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12339.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0071","submitter":"NVIDIA","system":"(72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":289712,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4023.777777777778,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0071","submitter":"NVIDIA","system":"(72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":167578,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2327.472222222222,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0071","submitter":"NVIDIA","system":"(72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":9921.02,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":137.79194444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0071","submitter":"NVIDIA","system":"(72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":14774.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":205.1986111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0071","submitter":"NVIDIA","system":"(72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":11614.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":161.3097222222222,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0072","submitter":"NVIDIA","system":"(72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":420659,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5842.486111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0072","submitter":"NVIDIA","system":"(72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":209328,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2907.3333333333335,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0072","submitter":"NVIDIA","system":"(72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":16104.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":223.66805555555555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0072","submitter":"NVIDIA","system":"(72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":12248.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":170.11666666666667,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":59622.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7452.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":101246,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12655.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":101611,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12701.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":59622.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7452.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101246,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12655.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":101611,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12701.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":770.668,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":96.3335,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1659.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":207.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0073","submitter":"Nebius","system":"Nebius B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1279.53,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":159.94125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":49683.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12420.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":49215.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12303.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":49683.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12420.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":49215.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12303.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":855.82,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":213.955,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0074","submitter":"Nebius","system":"Nebius GB200 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":596.108,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":149.027,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":23079.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2884.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":34812.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4351.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":34029.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4253.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":23079.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2884.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34812.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4351.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":34029.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4253.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":202.776,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":25.347,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":552.749,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":69.093625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0075","submitter":"Nebius","system":"Nebius H200 (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":295.36,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":36.92,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1615.85,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":201.98125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1243.67,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.45875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":145789,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18223.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128649,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16081.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":614344,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":76793,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0076","submitter":"Oracle","system":"BM.GPU.B200.8","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":43164.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5395.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0077","submitter":"Oracle","system":"BM.GPU.GB200.4","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":22272.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5568.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":12560.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":11801.5,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":458.848,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":365.409,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":12927.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0078","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1414.52,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":647861,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":80982.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":585168,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":73146,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":389330,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":48666.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":370068,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":46258.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":35091.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4386.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":33356.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4169.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":34876.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4359.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":33352.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4169.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":550.106,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":68.76325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":295.523,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":36.940375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":54117.6,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":6764.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":56917.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7114.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":57970.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7246.3,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":14539.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1817.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":13603.3,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1700.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":19.189,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.398625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":18.4233,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.3029125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0079","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D74H-7U (8x H200-SXM-141GB, TensorRT)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":33363.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4170.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Offline","result":263090,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":65772.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"dlrm-v2-99","benchmark_name":"DLRM-v2, 99% accuracy","scenario":"Server","result":202030,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":50507.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":159060,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":39765,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":154025,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":38506.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":6709.74,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1677.435,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":15057.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3764.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":13736.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3434.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":6712.97,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1678.2425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":15167.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3791.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":13730.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3432.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":21944.1,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":5486.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":25180.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6295.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":25753.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6438.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":6618.99,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1654.7475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":6600.5,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1650.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":8.35974,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.089935,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":7.82647,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.9566175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0080","submitter":"Quanta_Cloud_Technology","system":"D75E-4U_H200-NVL-141GBx4","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":15061.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3765.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":17633,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2204.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":32760.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4095.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31401.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3925.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":17633,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2204.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":32760.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4095.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31401.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3925.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":68218.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8527.3375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0081","submitter":"Quanta_Cloud_Technology","system":"D75T-7U_8xMI325X","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":59040.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7380.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0082","submitter":"RedHat","system":"IBM Cloud gx3 instance 1xL40S-PCIE-48GB ,vLLM","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1642.22,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1642.22,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0082","submitter":"RedHat","system":"IBM Cloud gx3 instance 1xL40S-PCIE-48GB ,vLLM","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":1207.14,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1207.14,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0083","submitter":"RedHat","system":"Dell PowerEdge XE8640 (1xH100-SXM-80GB, vLLM)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":5777.08,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5777.08,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0083","submitter":"RedHat","system":"Dell PowerEdge XE8640 (1xH100-SXM-80GB, vLLM)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":5103.99,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5103.99,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Offline","result":18928.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"dlrm-v2-99.9","benchmark_name":"DLRM-v2, 99.9% accuracy","scenario":"Server","result":18704.2,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1195.93,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":450.869,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Offline","result":716.683,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"retinanet","benchmark_name":"RetinaNet","scenario":"Server","result":579.509,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":25182.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0084","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2196.65,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0085","submitter":"Supermicro","system":"AS-8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":33468.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4183.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0085","submitter":"Supermicro","system":"AS-8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31646.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3955.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0085","submitter":"Supermicro","system":"AS-8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":33468.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4183.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0085","submitter":"Supermicro","system":"AS-8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31646.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3955.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":101516,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12689.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99186.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12398.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101865,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12733.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99182.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12397.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":122269,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":15283.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":142955,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17869.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":128701,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16087.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0086","submitter":"Supermicro","system":"AS-4126GS-NBR-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":44708.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5588.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":60140.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7517.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101859,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12732.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99195.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12399.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":751.385,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.923125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1641.34,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":205.1675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1240.44,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.055,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0087","submitter":"Supermicro","system":"SYS-422GA-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":42827.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5353.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":60131,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7516.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":102498,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12812.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99181.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12397.6375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":60119.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7514.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101696,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12712,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99179.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12397.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":750.939,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.867375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1650.51,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":206.31375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1244.69,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.58625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":141056,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17632,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":122331,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15291.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0088","submitter":"Supermicro","system":"SYS-422GS-NBRT-LCC (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":43615.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5451.9125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":57418,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7177.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101359,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12669.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99128.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12391.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":750.408,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":93.801,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1621.72,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":202.715,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1242.84,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":155.355,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0089","submitter":"Supermicro","system":"SYS-A21GE-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":44140.9,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5517.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0090","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":65320.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4082.50625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0090","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":57391.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3586.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0090","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":65320.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4082.50625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0090","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":57391.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3586.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0091","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost) and Supermicro AS-8125GS-TNMR2 (1x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3E (x16), AMD Instinct MI300X 192GB HBM3 (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":92158.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0091","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost) and Supermicro AS-8125GS-TNMR2 (1x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3E (x16), AMD Instinct MI300X 192GB HBM3 (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":80594.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0091","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost) and Supermicro AS-8125GS-TNMR2 (1x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3E (x16), AMD Instinct MI300X 192GB HBM3 (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":92158.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0091","submitter":"Supermicro_MangoBoost","system":"Supermicro AS-8126GS-TNMR (2x8x MI325X-256GB MangoBoost-LLMBoost) and Supermicro AS-8125GS-TNMR2 (1x8x MI300X-192GB MangoBoost-LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3E (x16), AMD Instinct MI300X 192GB HBM3 (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":80594.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0092","submitter":"The_Stage","system":"Nebius H100 (8x H100-SXM-80GB, TheStageAI)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":18.1007,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.2625875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0092","submitter":"The_Stage","system":"Nebius H100 (8x H100-SXM-80GB, TheStageAI)","accelerator":"NVIDIA H100-SXM-80GB","chip":"NVIDIA H100","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":17.8585,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.2323125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0093","submitter":"University-of-Florida","system":"NVIDIA DGX B200 (1x B200‑SXM‑180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":1,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":3.99057,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.99057,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0093","submitter":"University-of-Florida","system":"NVIDIA DGX B200 (1x B200‑SXM‑180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":1,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":3.47838,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.47838,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0094","submitter":"University-of-Florida","system":"NVIDIA DGX B200 (8x B200‑SXM‑180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Offline","result":31.935,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3.991875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0094","submitter":"University-of-Florida","system":"NVIDIA DGX B200 (8x B200‑SXM‑180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"stable-diffusion-xl","benchmark_name":"Stable Diffusion XL","scenario":"Server","result":28.4643,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":3.5580375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0094","submitter":"University-of-Florida","system":"NVIDIA DGX B200 (8x B200‑SXM‑180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":41918.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5239.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":17709.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2213.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":33762.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4220.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":30339.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3792.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":17709.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2213.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":33762.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4220.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0095","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":30339.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3792.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0096","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":70035.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8754.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":5.1,"result_id":"5.1-0096","submitter":"Vultr","system":"Supermicro AS -8126GS-TNMR (8x MI325X)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":56653.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7081.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2025-09-09","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v5.1/5ea4f62ef62536e6bf4d78a9b440fb9035ddfb4a/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v5.1-summary.json.gz","kept_sha256":"68c19b46cf5a87f535beae81154d61c3d981fe850ae3d09af530c769b6e0cc68","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":785522,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":1042110,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":1016380,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":785522,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":1042110,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0001","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x87)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 88 in all, 11 nodes of 8, x87 in the name","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":1016380,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":95004,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11875.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":82136.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10267.0125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":73608.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9201.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":103480,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12935,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":100282,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12535.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":73608.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9201.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":103480,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12935,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0002","submitter":"AMD","system":"smci355-ccs-aus-m09-17 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":100282,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12535.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0003","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x94)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 96 in all, 12 nodes of 8, x94 in the name","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1031070,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0003","submitter":"AMD","system":"AMD MI355X Custom Cluster","accelerator":"AMD Instinct MI355X 288GB HBM3e (x94)","chip":"AMD Instinct MI355X","chips":null,"chips_note":"the summary states 96 in all, 12 nodes of 8, x94 in the name","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":900054,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":6238.12,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":779.765,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27812.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3476.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":28038.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3504.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":34241,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4280.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":48791,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6098.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":48531.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6066.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0004","submitter":"ASUSTeK","system":"ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18338.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2292.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0005","submitter":"ASUSTeK","system":"ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":6262.63,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":782.82875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0005","submitter":"ASUSTeK","system":"ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27730.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3466.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0005","submitter":"ASUSTeK","system":"ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":27113.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3389.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0005","submitter":"ASUSTeK","system":"ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":48613.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6076.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0005","submitter":"ASUSTeK","system":"ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":47581,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5947.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":26005.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3250.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":110077,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13759.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100656,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12582,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":109625,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13703.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":102186,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12773.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":166745,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":20843.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0006","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":148067,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18508.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0007","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":806.105,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0007","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":258.538,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0007","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":14283.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0007","submitter":"Cisco","system":"C240M8-1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1422.99,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0008","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":30195.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3774.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0009","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18265.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2283.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69251.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8656.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":58553.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7319.1625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":111496,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13937,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":110655,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13831.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":115730,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14466.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":114504,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14313,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1971.11,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":246.38875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1445.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":180.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0573123,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0071640375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0010","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":21.425919695,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0011","submitter":"Cisco","system":"Cisco UCS  16xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":103633,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6477.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0011","submitter":"Cisco","system":"Cisco UCS  16xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":159932,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9995.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0011","submitter":"Cisco","system":"Cisco UCS  16xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":16,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":153022,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9563.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0012","submitter":"Cisco","system":"Cisco UCS  8xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":51616.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6452.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0012","submitter":"Cisco","system":"Cisco UCS  8xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":78936.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9867.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0012","submitter":"Cisco","system":"Cisco UCS  8xMI350X","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":76617.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9577.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0013","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":24216.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6054.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0013","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":25731.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6432.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0013","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":15338.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3834.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0014","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":13529.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3382.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0014","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":13029.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3257.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0014","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":24013.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6003.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0014","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":23423.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5855.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0014","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":9099.08,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2274.77,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0015","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (64x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":64,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":468665,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7322.890625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0015","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (64x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":64,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":311531,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4867.671875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0015","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (64x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":64,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":774539,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12102.171875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0015","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (64x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":64,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":743550,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11617.96875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0016","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (64x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":64,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":910468,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14226.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0016","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (64x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":64,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":956523,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14945.671875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0017","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":78564,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9820.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0017","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":44589.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5573.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0018","submitter":"Dell","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":807.675,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0018","submitter":"Dell","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":245.63,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0018","submitter":"Dell","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":14283.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0018","submitter":"Dell","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1414.26,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0019","submitter":"Dell","system":"PowerEdge XE9680 (8x MI300X)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27072.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3384.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0019","submitter":"Dell","system":"PowerEdge XE9680 (8x MI300X)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":24467.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3058.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0019","submitter":"Dell","system":"PowerEdge XE9680 (8x MI300X)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27072.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3384.0625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0019","submitter":"Dell","system":"PowerEdge XE9680 (8x MI300X)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":24467.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3058.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":91395.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11424.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":78603.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9825.4375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":70610.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8826.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":100008,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12501,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":96734.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12091.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":70610.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8826.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":100008,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12501,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0020","submitter":"Dell","system":"PowerEdge XE9785L (8x MI355X)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":96734.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12091.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":16343.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2042.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":32004,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4000.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":29084.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3635.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":16343.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2042.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":32004,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4000.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":29084.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3635.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":53193.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6649.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":45075.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5634.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0021","submitter":"Dell","system":"PowerEdge XE7740 (8x H200-NVL-141GB, TensorRT)","accelerator":"NVIDIA H200-NVL-141GB","chip":"NVIDIA H200 NVL","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":31969.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":3996.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0022","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27034.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3379.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0022","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":25524.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3190.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0022","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27034.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3379.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0022","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":25524.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3190.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0022","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18231.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2278.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":26829.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3353.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":25536.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3192.0125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":26829.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3353.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":25536.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3192.0125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":49036.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6129.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":45008.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5626.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0023","submitter":"Dell","system":"PowerEdge XE7745 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18433.5,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2304.1875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":66287.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8285.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":46510.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5813.825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":70724.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8840.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":108669,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13583.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":108392,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13549,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":70724.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8840.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":108669,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13583.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0024","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":108392,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13549,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0025","submitter":"Dell","system":"PowerEdge XE9780L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":45658.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5707.35,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0026","submitter":"Dell","system":"PowerEdge XE9780 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":58235.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7279.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0026","submitter":"Dell","system":"PowerEdge XE9780 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":40949.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5118.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0027","submitter":"Dell","system":"PowerEdge XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69021.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8627.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0027","submitter":"Dell","system":"PowerEdge XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":46513.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5814.1375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0031","submitter":"Dell_MangoBoost","system":"Dell_MangoBoost Heterogeneous Cluster (8x MI300X + 8x MI325X + 8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x8) and AMD Instinct MI325X 256GB HBM3e (x8) and AMD Instinct MI355X 288GB HBM3e (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":151843,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0031","submitter":"Dell_MangoBoost","system":"Dell_MangoBoost Heterogeneous Cluster (8x MI300X + 8x MI325X + 8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x8) and AMD Instinct MI325X 256GB HBM3e (x8) and AMD Instinct MI355X 288GB HBM3e (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":141521,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0031","submitter":"Dell_MangoBoost","system":"Dell_MangoBoost Heterogeneous Cluster (8x MI300X + 8x MI325X + 8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x8) and AMD Instinct MI325X 256GB HBM3e (x8) and AMD Instinct MI355X 288GB HBM3e (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":151843,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0031","submitter":"Dell_MangoBoost","system":"Dell_MangoBoost Heterogeneous Cluster (8x MI300X + 8x MI325X + 8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x8) and AMD Instinct MI325X 256GB HBM3e (x8) and AMD Instinct MI355X 288GB HBM3e (x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":141521,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0032","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":97535.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12191.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0032","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":93272.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11659.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0032","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":97535.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12191.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0032","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":93272.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11659.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0033","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785 (MI355X Power Cap 1000W)","accelerator":"AMD Instinct MI355X 288GB HBM3e (Power Cap 1000 W)","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":81870.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10233.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0033","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785 (MI355X Power Cap 1000W)","accelerator":"AMD Instinct MI355X 288GB HBM3e (Power Cap 1000 W)","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":76359.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9544.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0033","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785 (MI355X Power Cap 1000W)","accelerator":"AMD Instinct MI355X 288GB HBM3e (Power Cap 1000 W)","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":81870.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10233.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0033","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785 (MI355X Power Cap 1000W)","accelerator":"AMD Instinct MI355X 288GB HBM3e (Power Cap 1000 W)","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":76359.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9544.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":92332.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11541.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":80089.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10011.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":72547.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9068.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":100310,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12538.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":98157.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12269.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":72547.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9068.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":100310,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12538.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0037","submitter":"GigaComputing","system":"G893-ZX1-AAX4","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":98157.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12269.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0038","submitter":"GigaComputing","system":"DLB2-CB3","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":60.9525,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":15.238125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0038","submitter":"GigaComputing","system":"DLB2-CB3","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":41.6342,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":10.40855,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":4934.65,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":616.83125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69617.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8702.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":35552.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4444.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":107696,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13462,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":106397,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13299.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":65464.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8183.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":109343,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13667.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":104623,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13077.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":65464.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8183.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":109343,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13667.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":104623,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13077.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":764.655,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":95.581875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1947.91,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":243.48875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1340.51,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":167.56375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":128633,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16079.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":150707,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18838.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":141484,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17685.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0582849,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0072856125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":20.897894239,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0039","submitter":"GigaComputing","system":"G894-SD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":48116.6,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":6014.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":57116.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7139.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":41027.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5128.4875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":10447.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1305.9125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":10007,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1250.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1350.24,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":168.78,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":983.626,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":122.95325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0040","submitter":"Google","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":49327.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":6165.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0041","submitter":"Google","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":42.4736,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":10.6184,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0041","submitter":"Google","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":35.5625,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":8.890625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0041","submitter":"Google","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":22051.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5512.85,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0042","submitter":"Google","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":307874,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4276.027777777777,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0042","submitter":"Google","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":156781,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2177.5138888888887,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0043","submitter":"HPE","system":"HPE ProLiant DL145 Gen11 (1x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":2340.75,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2340.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0043","submitter":"HPE","system":"HPE ProLiant DL145 Gen11 (1x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":2081.34,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2081.34,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0044","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":41242.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4124.24,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0044","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":39234.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3923.41,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0044","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18889.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1888.94,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":19040.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1904.08,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":17737.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1773.7599999999998,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":36710.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3671.0800000000004,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":35590.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3559.0800000000004,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":36704.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3670.4199999999996,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":35522.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3552.2300000000005,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":53054.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5305.469999999999,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":60679.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6067.99,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0045","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":10,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":59632.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5963.2699999999995,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0046","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":33637.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4204.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0046","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":32151,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4018.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":15189.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1898.7375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":14258.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1782.3625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":29908.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3738.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":28727.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3590.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":29900.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3737.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":28717.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3589.6375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":44087.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5510.9125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":49580.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6197.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":48794.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6099.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0047","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18709,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2338.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":61644.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7705.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":103954,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12994.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":103635,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12954.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":61300.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7662.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":104572,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13071.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":103627,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12953.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":126143,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15767.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":152625,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19078.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":131270,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16408.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":126468,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15808.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":127209,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15901.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0048","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x NVIDIA B200 180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":46033.9,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5754.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":91846.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11480.825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":75354.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9419.3625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":70954.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8869.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":100019,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12502.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":95544.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11943.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":70954.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8869.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":100019,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12502.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0049","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":95544.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11943.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380 Gen12 (5x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":11665,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2333,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0050","submitter":"HPE","system":"HPE ProLiant Compute DL380 Gen12 (5x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":11352.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2270.42,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":6862.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3431.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":6662.33,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3331.165,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":10459.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5229.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":12433.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6216.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":11898.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5949.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":8198.25,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4099.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0051","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":7742.84,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3871.42,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":6901.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3450.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":6952.97,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3476.485,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":6893.36,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3446.68,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":6953.48,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3476.74,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":10589.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5294.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":12495.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6247.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":11899.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5949.85,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0052","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition-600W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":2,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":4712.66,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2356.33,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0053","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":19991.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4997.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0053","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":23265.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5816.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0053","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":22916.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5729.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0053","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Offline","result":14828.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3707.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0053","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"mixtral-8x7b","benchmark_name":"Mixtral 8x7B","scenario":"Server","result":13846.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3461.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":7297.05,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1824.2625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":6687.25,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1671.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":13361.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3340.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":12822.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3205.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":13200.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3300.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":12572.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3143.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":20118.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5029.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":23556.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5889.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":23227.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5806.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0054","submitter":"HPE","system":"HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition-450W, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":9228.36,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2307.09,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0057","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1229.56,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0057","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":471.541,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0057","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":24149.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0057","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2200.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0058","submitter":"Intel","system":"1-node-4x-BMG-B50","accelerator":"Intel(R) Arc Pro(R) B50","chip":"Intel Arc Pro B50","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1721.84,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":430.46,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0058","submitter":"Intel","system":"1-node-4x-BMG-B50","accelerator":"Intel(R) Arc Pro(R) B50","chip":"Intel Arc Pro B50","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":1187.13,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":296.7825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0058","submitter":"Intel","system":"1-node-4x-BMG-B50","accelerator":"Intel(R) Arc Pro(R) B50","chip":"Intel Arc Pro B50","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2112.35,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":528.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":841.038,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":210.2595,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":452.191,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":113.04775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":1697.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":424.415,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":1106.26,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":276.565,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":1697.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":424.415,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":1106.26,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":276.565,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":3347.43,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":836.8575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":3145.22,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":786.305,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0059","submitter":"Intel","system":"1-node-4x-BMG-B60","accelerator":"Intel(R) Arc Pro(R) B60","chip":"Intel Arc Pro B60","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":3995.63,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":998.9075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1601.91,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":400.4775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":884.238,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":221.0595,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":3270.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":817.665,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":2199.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":549.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":3270.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":817.665,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":2199.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":549.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":6762.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1690.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0060","submitter":"Intel","system":"1-node-4x-BMG-Pro-B60-Dual","accelerator":"MS-Intel Arc Pro B60 Dual 48G Turbo","chip":"Intel Arc Pro B60 Dual 48G","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":6293.89,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1573.4725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0061","submitter":"Krai","system":"HPE Cray XD670 (8x H200-SXM-141GB, KISS-V)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0181337,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0022667125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0061","submitter":"Krai","system":"HPE Cray XD670 (8x H200-SXM-141GB, KISS-V)","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":57.401991155,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0062","submitter":"Lambda_SIT","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":128750,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16093.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0062","submitter":"Lambda_SIT","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":160403,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":20050.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0062","submitter":"Lambda_SIT","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":130008,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16251,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0063","submitter":"Lambda_SIT","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":60220.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15055.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0063","submitter":"Lambda_SIT","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":53462.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13365.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0064","submitter":"Lenovo","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":775.982,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0064","submitter":"Lenovo","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":281.779,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0064","submitter":"Lenovo","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":13570.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0064","submitter":"Lenovo","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1411.42,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0065","submitter":"Lenovo","system":"ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":26889.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3361.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0065","submitter":"Lenovo","system":"ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":27160.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3395.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0065","submitter":"Lenovo","system":"ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":26889.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3361.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0065","submitter":"Lenovo","system":"ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":27160.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3395.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0065","submitter":"Lenovo","system":"ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18401.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2300.1375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0066","submitter":"Lenovo","system":"ThinkSystem SR680a V4 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":109141,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13642.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0066","submitter":"Lenovo","system":"ThinkSystem SR680a V4 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100687,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12585.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0066","submitter":"Lenovo","system":"ThinkSystem SR680a V4 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":47268.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5908.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0069","submitter":"MangoBoost","system":"MangoBoost Mi300X (8x MI300X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":27598.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3449.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0069","submitter":"MangoBoost","system":"MangoBoost Mi300X (8x MI300X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":25463.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3182.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0069","submitter":"MangoBoost","system":"MangoBoost Mi300X (8x MI300X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":27598.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3449.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0069","submitter":"MangoBoost","system":"MangoBoost Mi300X (8x MI300X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":25463.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3182.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0070","submitter":"MangoBoost","system":"MangoBoost Mi325X (8x MI325X, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":33880.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4235.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0070","submitter":"MangoBoost","system":"MangoBoost Mi325X (8x MI325X, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":31888.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3986.1125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0070","submitter":"MangoBoost","system":"MangoBoost Mi325X (8x MI325X, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":33880.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4235.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0070","submitter":"MangoBoost","system":"MangoBoost Mi325X (8x MI325X, LLMBoost)","accelerator":"AMD Instinct MI325X 256GB HBM3e","chip":"AMD Instinct MI325X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":31888.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3986.1125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":54730.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6841.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":78988.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9873.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":77377.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9672.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":54730.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6841.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":78988.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9873.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0071","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":77377.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9672.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0072","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":10547.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1318.425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0072","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":10007,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1250.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0072","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0464402,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.005805025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0072","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":25.458383354,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0072","submitter":"NVIDIA","system":"NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":47433.9,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5929.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":70325.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8790.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":42721.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5340.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":103961,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12995.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100328,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12541,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":112954,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14119.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":107318,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13414.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":112954,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14119.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":107318,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13414.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1951.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":243.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1460.25,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":182.53125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":165432,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":20679,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":141565,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17695.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0585234,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.007315425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0073","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":23.778912625,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0074","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":43.7662,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":10.94155,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0074","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":38.0313,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":9.507825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0074","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.0269817,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.006745425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0074","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":40.243719116,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":240318,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3337.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":486141,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6751.958333333333,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":336106,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4668.138888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":104637,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1453.2916666666667,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":99996.8,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1388.8444444444444,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":624929,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8679.569444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":879542,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12215.861111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":899218,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12489.138888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":754855,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10484.097222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":888054,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12334.083333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":810104,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11251.444444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":754855,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10484.097222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":888054,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12334.083333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":810104,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11251.444444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":14010.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":194.58472222222224,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":15461.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":214.74305555555554,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0075","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":14133.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":196.30416666666667,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0076","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":2494310,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8660.798611111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0076","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":1555110,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5399.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0077","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":61.2443,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":15.311075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0077","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":41.6351,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":10.408775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0077","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"Offline","result":0.036622,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0091555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0077","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"wan-2.2-t2v-a14b","benchmark_name":"Wan 2.2 text to video","scenario":"SingleStream","result":30.910508163,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":250634,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3481.027777777778,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":647076,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8987.166666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":400437,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5561.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":677199,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9405.541666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1042980,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14485.833333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1072250,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14892.361111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":814128,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11307.333333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":1126850,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15650.694444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":868278,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12059.416666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":814128,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11307.333333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":1126850,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15650.694444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":868278,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12059.416666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Interactive","result":18365.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":255.07500000000002,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":19511.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":270.99583333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0078","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":18627.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":258.71666666666664,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0079","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":1,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":15371.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15371.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0079","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":1,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":14973.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14973.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":60.6154,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":15.15385,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":43.1538,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":10.78845,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":673936,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9360.222222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":575580,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7994.166666666667,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1046150,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14529.861111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1096770,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15232.916666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0082","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":76346.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9543.325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0082","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":64510,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8063.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0083","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":58581.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7322.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0083","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":51692.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6461.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0083","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":13155.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1644.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0083","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":85921.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10740.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0083","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":87444.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10930.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69318.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8664.8625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":60413.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7551.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":106885,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13360.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100437,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12554.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":78.2775,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":9.7846875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0084","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":45.1513,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":5.6439125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0085","submitter":"Netweb_technologies","system":"Camarero PDI200A2HG-810 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":10737.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1342.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0085","submitter":"Netweb_technologies","system":"Camarero PDI200A2HG-810 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":10007,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1250.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0085","submitter":"Netweb_technologies","system":"Camarero PDI200A2HG-810 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":45083.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5635.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":91706.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11463.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":78186.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9773.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":72116.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9014.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":99874.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12484.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":98239.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12279.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":72116.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9014.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":99874.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12484.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0086","submitter":"ORACLE","system":"BM.GPU.MI355X.8","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":98239.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12279.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0087","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":777.671,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0087","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":284.359,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0087","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":13151.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0087","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1407.49,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0088","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75E-4U (8x RTX PRO6000 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":17838.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2229.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0089","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69263.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8657.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0089","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":35435.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4429.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0089","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Offline","result":1953.77,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":244.22125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0089","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-405b","benchmark_name":"Llama 3.1 405B","scenario":"Server","result":1484.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":185.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0090","submitter":"RedHat","system":"IBM Cloud gx3-48x240x2l40s instance,2xL40S-PCIE-48GB,vLLM 0.15.1","accelerator":"NVIDIA L40S","chip":"NVIDIA L40S","chips":2,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":3646.91,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1823.455,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0091","submitter":"RedHat","system":"LLM-D v0.5.0,Openshift 4.20.12,NVIDIA 8xB200-SXM-180GB","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":93070.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11633.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0091","submitter":"RedHat","system":"LLM-D v0.5.0,Openshift 4.20.12,NVIDIA 8xB200-SXM-180GB","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":71588.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8948.5125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0092","submitter":"RedHat","system":"LLM-D v0.5.0,OpenShift 4.21,IBM cloud gx3d-160x1792x8h200,NVIDIA 8xH200-SXM-141GB","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":28680,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3585,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0092","submitter":"RedHat","system":"LLM-D v0.5.0,OpenShift 4.21,IBM cloud gx3d-160x1792x8h200,NVIDIA 8xH200-SXM-141GB","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":24103.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3012.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0093","submitter":"RedHat","system":"IBM Cloud qx3d-160x1792x8h200,NVIDIA 8xH200-SXM-141GB,vLLM 0.15.1","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":36395.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":4549.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0094","submitter":"RedHat","system":"Dell B200,8xB200-SXM-180GB,RHEL 10.1,vLLM CentML:mlperf-inf-mm-q3vl-v6.0","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":79.0373,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":9.8796625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0094","submitter":"RedHat","system":"Dell B200,8xB200-SXM-180GB,RHEL 10.1,vLLM CentML:mlperf-inf-mm-q3vl-v6.0","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":67.8642,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":8.483025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0095","submitter":"RedHat","system":"IBM Cloud gx3d-160x1792x8h200,8xH200-SXM-141GB,RHEL 9.6,vLLM CentML:mlperf-inf-mm-q3vl-v6.0","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Offline","result":18.218,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2.27725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0095","submitter":"RedHat","system":"IBM Cloud gx3d-160x1792x8h200,8xH200-SXM-141GB,RHEL 9.6,vLLM CentML:mlperf-inf-mm-q3vl-v6.0","accelerator":"NVIDIA H200-SXM-141GB","chip":"NVIDIA H200","chips":8,"chips_note":null,"benchmark":"qwen3-vl-235b-a22b","benchmark_name":"Qwen3-VL-235B-A22B","scenario":"Server","result":11.0498,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.381225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0096","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1185.98,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0096","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":497.339,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0096","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":25192.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0096","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2204.09,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Interactive","result":71295.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8911.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":100754,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12594.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":97587,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12198.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":71295.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8911.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":100754,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12594.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0097","submitter":"Supermicro","system":"QuantaGrid D74A-7U","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":97587,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12198.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0098","submitter":"Supermicro","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":50562.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":6320.275,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0099","submitter":"Supermicro","system":"SYS-822GS-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":101595,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12699.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0099","submitter":"Supermicro","system":"SYS-822GS-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":99301.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12412.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0099","submitter":"Supermicro","system":"SYS-822GS-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101526,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12690.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0099","submitter":"Supermicro","system":"SYS-822GS-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":99304.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12413.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0099","submitter":"Supermicro","system":"SYS-822GS-NBRT (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":42389.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5298.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0100","submitter":"Supermicro_RedHat","system":"AS-8126GS-TNMR,vLLM","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Offline","result":80478.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10059.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6,"result_id":"6.0-0100","submitter":"Supermicro_RedHat","system":"AS-8126GS-TNMR,vLLM","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99","benchmark_name":"Llama 2 70B, 99% accuracy","scenario":"Server","result":76393.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9549.1375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-04-01","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.0/4d3916ac9cf474b679cdfcf492d43a0559418ad1/summary_results.json","kept_copy":"data/sources/mlperf-inference/kept/v6.0-summary.json.gz","kept_sha256":"52fa813a27834e8de38eca9fd381688df1cf9cd95020d2bd9e951f6cc16384ae","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0001","submitter":"AMD","system":"8xMI350X_2xEPYC_9575F","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":9297.78,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1162.2225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0001","submitter":"AMD","system":"8xMI350X_2xEPYC_9575F","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":10891.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1361.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0002","submitter":"AMD","system":"Quanta Grand Teton 1.5 MI350","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":20.603024726,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0002","submitter":"AMD","system":"Quanta Grand Teton 1.5 MI350","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0565013,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0070626625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":12198.8,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1524.85,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":13027.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1628.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":103275,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12909.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":106517,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13314.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":73191.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9148.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":16.124484197,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0731282,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.009141025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":113241,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14155.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0003","submitter":"AMD","system":"smci355-ccs-aus-m13-13 (AS -4126GS-NMR-LCC)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":121818,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15227.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0004","submitter":"AMD","system":"9x8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":964468,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13395.388888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0004","submitter":"AMD","system":"9x8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1039900,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14443.055555555555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":113455,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14181.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":113007,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14125.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":54568.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6821.075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69846.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8730.8625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":109949,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13743.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0005","submitter":"ASUSTeK","system":"XA NB3I-E12","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":109368,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13671,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0008","submitter":"Azure","system":"Azure GB200 NVL72 (288x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":1598510,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5550.381944444444,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0008","submitter":"Azure","system":"Azure GB200 NVL72 (288x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":2091190,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7261.076388888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0009","submitter":"Azure","system":"Azure GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":426796,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5927.722222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0009","submitter":"Azure","system":"Azure GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":542075,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7528.819444444444,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0010","submitter":"Azure","system":"Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":589391,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8185.986111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0010","submitter":"Azure","system":"Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":661143,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9182.541666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0010","submitter":"Azure","system":"Azure GB300 NVL72 (72x GB300-288GB_aarch64 TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":260098,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3612.472222222222,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0011","submitter":"Cisco","system":"Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":11426.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2285.2200000000003,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0011","submitter":"Cisco","system":"Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":12017.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2403.54,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0011","submitter":"Cisco","system":"Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":8494.68,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1698.9360000000001,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0011","submitter":"Cisco","system":"Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":5,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":4808.98,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":961.7959999999999,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":18186.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2273.275,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":18800.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2350.1125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":14316.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1789.525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0012","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":7777.66,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":972.2075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":49174.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6146.8375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":50453.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6306.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":46164.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5770.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":29316.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3664.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":29904,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3738,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":18991.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2373.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":14817.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1852.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0013","submitter":"Cisco","system":"Cisco UCS C845A M8 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":15508.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1938.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":173950,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":21743.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":171114,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":21389.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":129168,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16146,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":66171.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8271.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":70325.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8790.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":52221.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":6527.725,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":20.00421451803472,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0014","submitter":"Cisco","system":"Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0612749682980279,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0076593710372534875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":179597,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11224.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":227570,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14223.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":88915.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5557.21875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":137067,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8566.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":214774,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13423.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0015","submitter":"Cisco","system":"Cisco_UCS_B300x16_G200_4x2_RO","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":220527,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13782.9375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":161110,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":163292,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":121557,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":95104.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":99683.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":70354.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":73357.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0016","submitter":"Cisco","system":"Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR","accelerator":"NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":94249.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0017","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":9277.73,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2319.4325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0017","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":9518.66,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2379.665,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0017","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":6883.94,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1720.985,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0017","submitter":"Cisco","system":"Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell-32GB, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":3877.67,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":969.4175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0018","submitter":"Cisco","system":"Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":14333.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3583.325,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0018","submitter":"Cisco","system":"Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":14860.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3715.175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0018","submitter":"Cisco","system":"Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":6950.76,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1737.69,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0018","submitter":"Cisco","system":"Cisco UCS X210c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":7801.36,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1950.34,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0019","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":24498.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6124.65,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0019","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":24913.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6228.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0019","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":22447,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5611.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0019","submitter":"Cisco","system":"Cisco UCS X215c M8 - X580P PCIe Node (4x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":9236.94,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2309.235,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0020","submitter":"Cisco","system":"X410M8_1_node_4S_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":1030.45,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0020","submitter":"Cisco","system":"X410M8_1_node_4S_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":2068.63,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0020","submitter":"Cisco","system":"X410M8_1_node_4S_GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2970.57,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":102398,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12799.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":102703,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12837.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":56121.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7015.1625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":58913.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7364.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":90204.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11275.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0021","submitter":"CoreWeave","system":"CoreWeave B200 SXM (8x B200-SXM-180GB)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":91487.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11435.925,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0022","submitter":"CoreWeave","system":"CoreWeave B300 SXM (8x B300-SXM-270GB)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":114576,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14322,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0022","submitter":"CoreWeave","system":"CoreWeave B300 SXM (8x B300-SXM-270GB)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":115530,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14441.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0022","submitter":"CoreWeave","system":"CoreWeave B300 SXM (8x B300-SXM-270GB)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":111247,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13905.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0022","submitter":"CoreWeave","system":"CoreWeave B300 SXM (8x B300-SXM-270GB)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":112840,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14105,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":910621,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12647.513888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":894002,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12416.694444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":419778,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5830.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":529991,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7360.986111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":901058,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12514.694444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0023","submitter":"CoreWeave","system":"CoreWeave GB200 NVL72 (72x GB200-186GB_aarch64)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":911566,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12660.638888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0024","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (4x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":64.13202030676248,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.03300507669062,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0024","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (4x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":72.59800294104302,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":18.149500735260755,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":944902,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13123.638888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":1136100,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15779.166666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":1195.590758633778,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.60542720324692,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":1135.25001579439,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":15.76736133047764,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1160490,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16117.916666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0025","submitter":"CoreWeave","system":"CoreWeave GB300 NVL72 (72x GB300-288GB_aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1197710,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16634.86111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0026","submitter":"Crusoe","system":"Crusoe Cloud 64x8 MI355X (512 GPU) — DeepSeek-R1 / SGLang","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":512,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":2405310,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4697.87109375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0026","submitter":"Crusoe","system":"Crusoe Cloud 64x8 MI355X (512 GPU) — DeepSeek-R1 / SGLang","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":512,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":2901950,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5667.87109375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0027","submitter":"Crusoe","system":"Crusoe Cloud 64x8 MI355X (512 GPU) — gpt-oss-120b / vLLM","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":512,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":5392930,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10533.06640625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0027","submitter":"Crusoe","system":"Crusoe Cloud 64x8 MI355X (512 GPU) — gpt-oss-120b / vLLM","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":512,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":5749440,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11229.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0028","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":16,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":198.9444574056901,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12.43402858785563,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0028","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":16,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":230.088792881784,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":14.3805495551115,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0028","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (16x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":16,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":39.85079133425917,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":2.490674458391198,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0029","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":49.6962883240102,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12.42407208100255,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0029","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":58.39587523885307,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":14.598968809713268,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0029","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, Dynamo)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":3.994056744750478,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":0.9985141861876194,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0030","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":50613.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12653.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0030","submitter":"Crusoe","system":"NVIDIA GB200 NVL72 (4x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":50912.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12728.225,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0031","submitter":"Dell","system":"PowerEdge R770AP_128C 24x96GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":21835.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0031","submitter":"Dell","system":"PowerEdge R770AP_128C 24x96GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2275.93,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0032","submitter":"Dell","system":"PowerEdge R770_86C 16x64GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":516.012,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0032","submitter":"Dell","system":"PowerEdge R770_86C 16x64GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1227.99,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0032","submitter":"Dell","system":"PowerEdge R770_86C 16x64GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":16042.8,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0032","submitter":"Dell","system":"PowerEdge R770_86C 16x64GB (6400 MT/s)","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1539.49,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0033","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":28973.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3621.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0033","submitter":"Dell","system":"PowerEdge XE7740 (8x RTXPro6000-PCIe-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":29796.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3724.5875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0034","submitter":"Dell","system":"PowerEdge XE7745 (8x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":3012.33,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":376.54125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0034","submitter":"Dell","system":"PowerEdge XE7745 (8x NVIDIA RTX PRO 4500 Blackwell, TensorRT)","accelerator":"NVIDIA RTX PRO 4500 Blackwell","chip":"NVIDIA RTX PRO 4500 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":4374.06,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":546.7575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0035","submitter":"Dell","system":"Dell PowerEdge XE9712","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":10.58500054065278,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0035","submitter":"Dell","system":"Dell PowerEdge XE9712","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.6455254875195523,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.008965631771104893,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0036","submitter":"Dell","system":"Dell XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"Nvidia B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":96180.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12022.575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0036","submitter":"Dell","system":"Dell XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"Nvidia B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":104006,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13000.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0036","submitter":"Dell","system":"Dell XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"Nvidia B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":103332,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12916.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0036","submitter":"Dell","system":"Dell XE9780 (8x B300-SXM-270GB, TensorRT)","accelerator":"Nvidia B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":109257,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13657.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":148563,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18570.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":158458,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19807.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":130540,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16317.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":103932,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12991.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":103792,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12974,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":73238.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9154.7625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":110188,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13773.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0037","submitter":"Dell","system":"XE9785L_8xMI355X","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":117804,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14725.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0038","submitter":"Dell","system":"Dell XE9785L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106898,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13362.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0038","submitter":"Dell","system":"Dell XE9785L (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":113101,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14137.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Server","result":4503.53,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":562.94125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"dlrm-v3","benchmark_name":"DLRM-v3","scenario":"Offline","result":4837.63,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":604.70375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":72305.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9038.1375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":73643,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9205.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":62723.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7840.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":41830,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5228.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":42286.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5285.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":22860.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":2857.5625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":41.83792310800001,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0345875,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0043234375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":40867.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5108.4875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0040","submitter":"Dell_AMD","system":"XE7745_8xMI350P_2xEPYC_9455","accelerator":"AMD Instinct MI350P","chip":"AMD Instinct MI350P","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":48971.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6121.3875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0041","submitter":"Dell_MangoBoost","system":"Heterogeneous Four-Cluster (2x8x MI300X + 2x8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x2x8) and AMD Instinct MI355X 288GB HBM3e (x2x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":253501,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0041","submitter":"Dell_MangoBoost","system":"Heterogeneous Four-Cluster (2x8x MI300X + 2x8x MI355X, LLMBoost)","accelerator":"AMD Instinct MI300X 192GB HBM3 (x2x8) and AMD Instinct MI355X 288GB HBM3e (x2x8)","chip":null,"chips":null,"chips_note":"the system has more than one kind of chip","benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":285454,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0042","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":109065,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13633.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0042","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":115294,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14411.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0042","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":28265.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3533.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0043","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785L","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":110153,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13769.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0043","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785L","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":116953,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14619.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0043","submitter":"Dell_MangoBoost","system":"PowerEdge XE9785L","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":29585.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3698.1875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0044","submitter":"Fujitsu","system":"PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":27340.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3417.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0044","submitter":"Fujitsu","system":"PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":26626.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3328.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":114832,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14354,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":115034,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14379.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":70495.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8811.9625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":72445.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9055.6375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":19.07940438844444,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0629152672355444,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.00786440840444305,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":111.7532977914224,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":13.9691622239278,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":131.2852847183584,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.4106605897948,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":12.48543213240489,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.5606790165506113,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":111218,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13902.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0045","submitter":"GigaComputing","system":"G894-AD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":111625,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13953.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":114578,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14322.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":114826,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14353.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":70771.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8846.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":72327.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9040.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":18.05179867184028,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0625123079679506,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.007814038495993825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":115.7839399384009,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":14.472992492300113,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":132.0547880616454,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.506848507705676,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":12.88177974080564,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.610222467600705,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":110592,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13824,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0046","submitter":"GigaComputing","system":"G894-ZD3-AAX7","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":111541,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13942.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0047","submitter":"Google","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":388590,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5397.083333333333,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0047","submitter":"Google","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":510781,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7094.180555555556,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0047","submitter":"Google","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":232160,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3224.4444444444443,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0048","submitter":"HPE","system":"NVIDIA GB200 NVL72 by HPE (8x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100714,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12589.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0048","submitter":"HPE","system":"NVIDIA GB200 NVL72 by HPE (8x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":101453,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12681.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":49523.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6190.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":50965.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6370.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":39986.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4998.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":28643.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3580.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":29284.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3660.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":19278.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":2409.7875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":14581.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1822.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0049","submitter":"HPE","system":"HPE ProLiant Compute DL380a Gen12 (8x RTX Pro 6000 Blackwell Server Edition 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":15256,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1907,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":148118,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18514.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":157694,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19711.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":130613,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16326.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":101979,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12747.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":102991,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12873.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":72977.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9122.15,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":16.397060151,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0707035,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0088379375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":108836,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13604.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0050","submitter":"HPE","system":"HPE ProLiant Compute XD685 (8x AMD MI355X-288GB, vLLM)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":117355,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14669.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0051","submitter":"HPE","system":"HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":80536.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5033.55625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0051","submitter":"HPE","system":"HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":16,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":136854,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8553.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0052","submitter":"HPE","system":"HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":113685,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14210.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0052","submitter":"HPE","system":"HPE Compute XD690 (8x NVIDIA B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":114569,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14321.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0059","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":1139.51,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0059","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1916.74,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0059","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":24573.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0059","submitter":"Intel","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2319.02,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0060","submitter":"Intel","system":"1-node-2S-Xeon_6787P","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"end to end question-answering","benchmark_name":"End-to-end question answering","scenario":"Offline","result":0.464076,"units":"Tasks/s","status":"Result verified by MLCommons Association.","per_chip":0.116019,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0060","submitter":"Intel","system":"1-node-2S-Xeon_6787P","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"end to end vector-database","benchmark_name":"End-to-end vector database","scenario":"Offline","result":22.6141,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5.653525,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":4173.25,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1043.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":4613.28,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1153.32,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":1745.48,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":436.37,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":2464.98,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":616.245,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":5539.77,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1384.9425,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1296.87,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":324.2175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0061","submitter":"Intel","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1956.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":489.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106617,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13327.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":109933,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13741.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":57432,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7179,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":70269.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8783.7,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100299,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12537.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0062","submitter":"Inventec","system":"Inventec P5800G7(LC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":106280,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13285,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106645,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13330.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":109755,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13719.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":57517.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7189.675,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69062.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8632.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100300,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12537.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0063","submitter":"Inventec","system":"Inventec P9000G7(AC) (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":105617,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13202.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0064","submitter":"Lambda","system":"Lambda Cloud 8x B200-SXM-180GB","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":69.41536196217709,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":8.676920245272136,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0064","submitter":"Lambda","system":"Lambda Cloud 8x B200-SXM-180GB","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":101.5607239508195,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":12.695090493852437,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0065","submitter":"Lambda","system":"Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":64.62448277938189,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.156120694845473,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0065","submitter":"Lambda","system":"Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":71.1217334486934,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":17.78043336217335,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0065","submitter":"Lambda","system":"Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":58195.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14548.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0065","submitter":"Lambda","system":"Lambda 4x GB300-288GB (GB300 NVL4, aarch64)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":65511.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16377.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0066","submitter":"MangoBoost","system":"MangoBoost 2x8 MI355X (16x MI355X, LLMBoost P/D Disaggregated Interactive Engine)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":16,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":228488,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14280.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0066","submitter":"MangoBoost","system":"MangoBoost 2x8 MI355X (16x MI355X, LLMBoost P/D Disaggregated Interactive Engine)","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":16,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":59297.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3706.0875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0067","submitter":"MangoBoost","system":"8xMI300X_2xEPYC_9534","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":24863,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3107.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0067","submitter":"MangoBoost","system":"8xMI300X_2xEPYC_9534","accelerator":"AMD Instinct MI300X 192GB HBM3","chip":"AMD Instinct MI300X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":30198.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3774.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":121706,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15213.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":129943,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16242.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":104368,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13046,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":85004.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10625.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":84016.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10502.05,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":56673,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7084.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":87549.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":10943.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0068","submitter":"MiTAC","system":"G8825Z5","accelerator":"AMD Instinct MI350X 288GB HBM3e","chip":"AMD Instinct MI350X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":97019.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12127.4625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":151116,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18889.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":159278,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19909.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":136999,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":17124.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":104541,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13067.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":104904,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13113,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":74385.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9298.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":108861,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13607.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0069","submitter":"MiTAC","system":"G4826","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":118042,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14755.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0070","submitter":"MiTAC","system":"MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":29203.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3650.4125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0070","submitter":"MiTAC","system":"MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":30034.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3754.275,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106949,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13368.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":109277,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13659.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":59867.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7483.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":70132.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8766.5375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":17.97718330692361,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0621886628274087,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.007773582853426088,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":104.5781916389203,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":13.072273954865038,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":129.3583917213701,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.169798965171264,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":3.994428963207542,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":0.4993036204009427,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100630,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12578.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0071","submitter":"NVIDIA","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":106637,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13329.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":395940,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":5499.166666666667,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":523790,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7274.861111111111,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":240274,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3337.1388888888887,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":5.731549943319444,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.4960942571652712,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.006890198016184322,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":899005,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12486.180555555555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":883761,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12274.458333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0072","submitter":"NVIDIA","system":"NVIDIA GB200 NVL72 (72x GB200-186GB_aarch64, TensorRT)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":625205,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8683.402777777777,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0073","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":2028030,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7041.770833333333,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0073","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":288,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":2705130,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9392.8125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":596944,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8290.888888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":679740,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9440.833333333334,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":253506,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3520.9166666666665,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":5.807651162902777,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.6546111599325777,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.009091821665730245,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":1210.48823482974,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.8123365948575,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":1304.985262579748,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":18.12479531360761,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":349.2921988473666,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":4.851280539546758,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1120710,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15565.416666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1124500,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15618.055555555555,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0074","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Interactive","result":669305,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9295.902777777777,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":127911,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15988.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":131464,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16433,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":67578.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8447.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":75918.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9489.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":128173,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16021.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0075","submitter":"NVIDIA","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":131619,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16452.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0077","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":55869,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6983.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0077","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":56929.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7116.2375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0077","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":89856.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11232.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0077","submitter":"Nebius","system":"Nebius B200 n1 (8x B200-SXM-180GB, TensorRT)","accelerator":"NVIDIA B200-SXM-180GB","chip":"NVIDIA B200","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":88491,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":11061.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":63909.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7988.6625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69655.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8706.8875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":112.2858237129371,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":14.035727964117138,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":131.4771630075575,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.434645375944687,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":108124,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13515.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0078","submitter":"Nebius","system":"Nebius B300 n1 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":108529,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13566.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0079","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":64.61995222586359,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.154988056465896,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0079","submitter":"Nebius","system":"Nebius GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":73.18605442300793,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":18.296513605751983,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":603023,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8375.319444444445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":689961,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9582.791666666666,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1122490,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15590.138888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0080","submitter":"Nebius","system":"Nebius GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1186750,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16482.63888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":127921,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15990.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0081","submitter":"Nebius","system":"Nebius GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":132236,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16529.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0082","submitter":"Nebius","system":"Nebius RTX PRO 6000 n1 (8x RTX-PRO-6000-PCIE-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":14784.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1848.0375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0082","submitter":"Nebius","system":"Nebius RTX PRO 6000 n1 (8x RTX-PRO-6000-PCIE-96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":15738.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1967.3125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0083","submitter":"Oracle","system":"BM.Standard4.Ax.120","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":324.995,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0083","submitter":"Oracle","system":"BM.Standard4.Ax.120","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":853.14,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0083","submitter":"Oracle","system":"BM.Standard4.Ax.120","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1099.64,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":148594,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18574.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":144321,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18040.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Interactive","result":130625,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16328.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":100392,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12549,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":101553,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12694.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":70421.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8802.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":16.334597792,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0706215,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0088276875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":109083,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13635.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0084","submitter":"Oracle","system":"8xMI355X_2xEPYC_9575F","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":115567,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14445.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106696,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13337,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":108026,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13503.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":60745.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7593.1875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":70354,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8794.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":124.3859737873004,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":15.54824672341255,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":3.9944621629557,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":0.4993077703694625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100608,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12576,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0085","submitter":"Oracle","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":109931,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13741.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":58991.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14747.975,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":66001.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16500.475,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":30.653359842875,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0138372212111634,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.00345930530279085,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":64.60782589940354,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":16.151956474850884,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":72.78618281809769,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":18.19654570452442,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Interactive","result":4.992209807647115,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":1.2480524519117788,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":59996,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14999,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0086","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":60745.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15186.45,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":586852,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8150.722222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":666613,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9258.513888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Interactive","result":235807,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":3275.097222222222,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":5.895425553486111,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.3997628673298144,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.005552262046247422,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":1146.388942892591,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":15.922068651285988,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":1297.490941230584,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":18.020707517091445,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1092610,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15175.138888888889,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0087","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (72x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":72,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":1097620,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15244.722222222223,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0088","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":127984,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15998,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0088","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":132253,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":16531.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0088","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":68138,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8517.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0088","submitter":"Oracle","system":"NVIDIA GB300 NVL72 (8x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":77031.7,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9628.9625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0089","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":516.012,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0089","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1170.51,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0089","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":15633.3,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0089","submitter":"Quanta_Cloud_Technology","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1548.1,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0090","submitter":"Quanta_Cloud_Technology","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":4122.73,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1030.6825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0090","submitter":"Quanta_Cloud_Technology","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":4977.67,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1244.4175,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0090","submitter":"Quanta_Cloud_Technology","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":5401.23,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1350.3075,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106285,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13285.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":113985,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14248.125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":59622.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7452.775,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69841.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8730.2,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":45408.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":5676.025,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":104.8118239643209,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":13.101477995540112,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":131.7949949301423,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.47437436626779,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":100183,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12522.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0091","submitter":"Quanta_Cloud_Technology","system":"QuantaGrid D75H-10U (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":108293,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13536.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0092","submitter":"RedHat_Intel","system":"1-node-2S-GNR_96C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":644.788,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0092","submitter":"RedHat_Intel","system":"1-node-2S-GNR_96C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1507.21,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0092","submitter":"RedHat_Intel","system":"1-node-2S-GNR_96C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1966.11,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0093","submitter":"RedHat_Supermicro","system":"NVIDIA GB200 NVL4 (4x GB200-192GB_aarch64, vLLM)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":50.68378958948414,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":12.670947397371036,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0093","submitter":"RedHat_Supermicro","system":"NVIDIA GB200 NVL4 (4x GB200-192GB_aarch64, vLLM)","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":58.07798401246531,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":14.519496003116327,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0094","submitter":"RedHat_Supermicro","system":"SMC Lab OpenShift GB200-NVL4 4xGB200-186GB","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":51095.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12773.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0094","submitter":"RedHat_Supermicro","system":"SMC Lab OpenShift GB200-NVL4 4xGB200-186GB","accelerator":"NVIDIA GB200","chip":"NVIDIA GB200","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":61205.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15301.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0095","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":1101.19,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0095","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1833.97,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0095","submitter":"Supermicro","system":"1-node-2S-GNR_128C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2248.15,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0096","submitter":"Supermicro","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":516.059,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0096","submitter":"Supermicro","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":1189.89,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0096","submitter":"Supermicro","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"rgat","benchmark_name":"R-GAT","scenario":"Offline","result":13273.7,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0096","submitter":"Supermicro","system":"1-node-2S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":1477.4,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0097","submitter":"Supermicro","system":"1-node-4S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":966.241,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0097","submitter":"Supermicro","system":"1-node-4S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":2063.24,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0097","submitter":"Supermicro","system":"1-node-4S-GNR_86C","accelerator":null,"chip":null,"chips":null,"chips_note":"the system has no accelerator","benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":2881.28,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":4208.73,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1052.1825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":4596.62,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":1149.155,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":1759.95,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":439.9875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":2484.14,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":621.035,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":5854.85,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":1463.7125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":1351.13,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":337.7825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0098","submitter":"Supermicro","system":"1-node-4x-BMG-B70","accelerator":"Intel(R) Arc(TM) Pro B70","chip":"Intel Arc Pro B70","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":2063.93,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":515.9825,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":148710,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":18588.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":159060,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19882.5,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":103032,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12879,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":102243,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":12780.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Interactive","result":73181.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":9147.6875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":109184,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13648,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0099","submitter":"Supermicro","system":"AS-4126GS-NMR-LCC","accelerator":"AMD Instinct MI355X 288GB HBM3e","chip":"AMD Instinct MI355X","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":117005,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14625.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":106872,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13359,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":110802,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13850.25,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":61164.9,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":7645.6125,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69830,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8728.75,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":107288,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13411,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0100","submitter":"Supermicro","system":"AS-8126GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":108727,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13590.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":109848,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13731,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":111247,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13905.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":66754.3,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8344.2875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":69104,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":8638,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Server","result":111.7738365067543,"units":"Queries/s","status":"Result verified by MLCommons Association.","per_chip":13.971729563344287,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"VLM","benchmark_name":"Vision-language model (VLM)","scenario":"Offline","result":128.9061241200454,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":16.113265515005676,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":107205,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13400.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0101","submitter":"Supermicro","system":"SYS-822GS-NB3RT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":107851,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13481.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0102","submitter":"TTA","system":"ManyCoreSoft DeepGadget dg5W (4x NVIDIA RTX PRO 6000 Blackwell 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":24510.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6127.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0102","submitter":"TTA","system":"ManyCoreSoft DeepGadget dg5W (4x NVIDIA RTX PRO 6000 Blackwell 96GB, TensorRT)","accelerator":"NVIDIA RTX PRO 6000 Blackwell Server Edition","chip":"NVIDIA RTX PRO 6000 Blackwell","chips":4,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":24511.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":6127.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":20420.4,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":20420.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":21266.1,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":21266.1,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Server","result":14495.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14495.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"llama2-70b-99.9","benchmark_name":"Llama 2 70B, 99.9% accuracy","scenario":"Offline","result":14636.8,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14636.8,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":7058.95,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":7058.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":13854,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":13854,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0103","submitter":"VibeHPC","system":"B300-SXM-270GBx1_TRT","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":1,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":14130.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":14130.6,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Server","result":156867,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":19608.375,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"llama3.1-8b","benchmark_name":"Llama 3.1 8B","scenario":"Offline","result":161767,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":20220.875,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Server","result":33415.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4176.9,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"deepseek-r1","benchmark_name":"DeepSeek-R1","scenario":"Offline","result":36639.6,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":4579.95,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"whisper","benchmark_name":"Whisper","scenario":"Offline","result":53267.2,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":6658.4,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"SingleStream","result":21.279754716,"units":"Latency (s)","status":"Result verified by MLCommons Association.","per_chip":null,"per_chip_status":null,"published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0104","submitter":"VibeHPC","system":"NVIDIA DGX B300 (8x B300-SXM-270GB, TensorRT)","accelerator":"NVIDIA B300-SXM-270GB","chip":"NVIDIA B300","chips":8,"chips_note":null,"benchmark":"text_to_video","benchmark_name":"Text to video","scenario":"Offline","result":0.0783824,"units":"Samples/s","status":"Result verified by MLCommons Association.","per_chip":0.0097978,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0105","submitter":"Wiwynn","system":"OneW GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Server","result":60466.2,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15116.55,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."},{"round":6.1,"result_id":"6.1-0105","submitter":"Wiwynn","system":"OneW GB300 NVL72 (4x GB300-288GB_aarch64, TensorRT)","accelerator":"NVIDIA GB300","chip":"NVIDIA GB300","chips":4,"chips_note":null,"benchmark":"gpt-oss-120b","benchmark_name":"GPT-OSS-120B","scenario":"Offline","result":63990.5,"units":"Tokens/s","status":"Result verified by MLCommons Association.","per_chip":15997.625,"per_chip_status":"A figure per chip is this site’s division of a result by the system’s count of accelerators. It is not a primary metric of MLPerf® Inference, and it is not verified by MLCommons Association.","published":"2026-09-16","source_url":"https://raw.githubusercontent.com/mlcommons/inference_results_v6.1/4bb63cd28eb136be221bc801a9e38c66c8a7553f/summary.xlsx","kept_copy":"data/sources/mlperf-inference/kept/v6.1-summary.xlsx.gz","kept_sha256":"226adca4c8cf18b64d56d8fe27ead58b86933dcb157a3d3f55bb9ff5bd164972","page":"https://jesseahawkins.com/chips/inference/","license":"MLCommons results repository, Apache License 2.0","trademark":"The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information."}]}