Frontier · open-weight downloads

Alibaba drew 45% of the Frontier labs’ open-model downloads

Alibaba: 251 million downloads in the 30 days to 3 Nov 2026, of 555 million for 13 labs.

The Hugging Face Hub shows each model’s downloads for the last 30 days only. This site reads every model of each lab’s Hub accounts once a week and keeps each answer with its sha256. One read so far, on 3 Nov 2026. A read is added each week.

downloads.csv downloads-labs.csv Hub read 3 Nov 2026

Embed

The frame credits the Hugging Face Hub and links back to its page. Preview it.

Cite
Download
Post on X

The labs, by downloads in 30 days

555 million together, to 3 Nov 2026
  1. Alibaba 496 models 251M
  2. Google 1,136 models 90M
  3. Meta 2,432 models 72M
  4. OpenAI 39 models 54M
  5. Microsoft 554 models 39M
  6. DeepSeek 106 models 24M
  7. Z.ai 154 models 17M
  8. Tencent 176 models 4.6M
  9. Moonshot AI 19 models 2.2M
  10. StepFun 54 models 535K
  11. Cohere 55 models 522K
  12. Xiaomi 30 models 359K
  13. xAI 2 models 10K

A lab’s count is the sum over every model its own Hub accounts publish. A copy of a model that another user posts counts for that user. Anthropic’s Hub account, Anthropic, holds no model. Recraft publishes no model weights. The Hub account recraft-ai holds no model on 2026-11-03.

Each lab’s totals

the rise since the last read shows from the second read
Lab Hub accounts Models Downloads, 30 days Downloads, all time Likes
Alibaba Qwen, Wan-AI 496 250,952,321 3,987,479,869 161,788
Google google 1,136 89,769,686 2,922,119,230 85,937
Meta meta-llama, facebook 2,432 71,688,366 3,643,442,105 117,878
OpenAI openai 39 54,365,438 3,172,887,317 32,582
Microsoft microsoft 554 38,869,730 2,346,873,784 59,816
DeepSeek deepseek-ai 106 23,823,933 417,019,571 87,253
Z.ai zai-org 154 17,065,617 185,232,266 47,280
Tencent tencent 176 4,579,733 29,000,640 33,902
Moonshot AI moonshotai 19 2,184,392 55,143,215 26,879
StepFun stepfun-ai 54 534,519 15,681,359 7,854
Cohere CohereLabs 55 522,473 12,395,484 11,419
Xiaomi XiaomiMiMo 30 359,424 7,872,799 5,841
xAI xai-org 2 9,918 314,229 3,690

Each lab, read by read

downloads in the 30 days to each read, and the lab’s top 10 models

One read so far, so each line is one point. A point is added each week. A week that is not read cannot be filled in later.

The top 100 models on the whole Hub

37 of them from a Frontier lab’s account
Rank Model Task Downloads, 30 days All time Likes
1 sentence-transformers/all-MiniLM-L6-v2 sentence-similarity 182,509,097 4,200,257,364 6,059
2 cross-encoder/ms-marco-MiniLM-L6-v2 text-ranking 66,901,235 679,066,731 439
3 BAAI/bge-small-en-v1.5 feature-extraction 49,836,681 545,450,747 662
4 Comfy-Org/MiniMax-H3 — 31,098,855 79,068,221 2,491
5 BAAI/bge-m3 sentence-similarity 30,252,576 355,857,436 3,961
6 openai-community/gpt2 text-generation 29,307,840 962,536,600 3,913
7 google-bert/bert-base-uncased fill-mask 29,234,461 3,281,853,805 3,060
8 sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 sentence-similarity 28,384,395 665,513,772 1,497
9 argmaxinc/whisperkit-coreml automatic-speech-recognition 27,332,146 107,782,889 260
10 autogluon/chronos-2 time-series-forecasting 23,513,356 97,976,455 94
11 google-t5/t5-small translation 23,416,949 301,650,809 687
12 Qwen/Qwen3-0.6BAlibaba text-generation 23,175,586 258,109,736 1,869
13 sentence-transformers/all-mpnet-base-v2 sentence-similarity 20,046,382 1,369,448,988 1,403
14 google/electra-base-discriminatorGoogle — 17,601,325 970,264,158 218
15 openai/clip-vit-base-patch32OpenAI zero-shot-image-classification 17,564,725 710,757,874 1,272
16 jonatasgrosman/wav2vec2-large-xlsr-53-japanese automatic-speech-recognition 16,178,890 123,614,599 118
17 google/vit-base-patch16-224Google image-classification 15,440,331 204,657,718 1,048
18 timm/mobilenetv3_small_100.lamb_in1k image-classification 15,087,982 1,188,391,071 156
19 BAAI/bge-reranker-v2-m3 text-classification 14,444,120 162,879,247 1,286
20 Qwen/Qwen3-14BAlibaba text-generation 12,988,834 43,111,159 532
21 nomic-ai/nomic-embed-text-v1.5 sentence-similarity 11,853,539 175,003,787 988
22 hexgrad/Kokoro-82M text-to-speech 11,366,340 138,899,285 7,332
23 Comfy-Org/Qwen-Image-2.1 text-to-image 10,607,935 16,860,819 1,255
24 prithivMLmods/Gender-Classifier-Mini image-classification 10,265,515 10,848,350 19
25 amazon/chronos-2 time-series-forecasting 9,968,412 192,793,310 565
26 Comfy-Org/Z-Image-Turbo — 9,521,322 53,579,153 1,011
27 Comfy-Org/Wan-2.2 image-to-video 9,326,230 98,314,213 992
28 intfloat/multilingual-e5-small sentence-similarity 9,281,664 128,517,956 480
29 BAAI/bge-large-en-v1.5 feature-extraction 9,177,978 178,109,902 781
30 sentence-transformers/paraphrase-multilingual-mpnet-base-v2 sentence-similarity 8,985,672 147,128,258 554
31 Qwen/Qwen3-Embedding-0.6BAlibaba feature-extraction 8,874,546 100,857,936 1,322
32 Bingsu/adetailer — 8,771,059 361,951,495 824
33 Qwen/Qwen3-4BAlibaba text-generation 8,417,695 93,821,375 780
34 FacebookAI/xlm-roberta-base fill-mask 8,382,998 736,802,481 967
35 facebook/contrieverMeta — 8,280,816 197,590,453 132
36 google/gemma-4-31B-itGoogle image-text-to-text 8,147,734 72,426,611 4,276
37 BAAI/bge-base-en-v1.5 feature-extraction 8,054,219 584,427,390 540
38 audio-cpp/audio.cpp-gguf text-to-speech 8,003,097 15,986,165 297
39 Qwen/Qwen3.8-27BAlibaba image-text-to-text 7,915,637 20,706,092 19,028
40 Qwen/Qwen3-8BAlibaba text-generation 7,540,663 136,299,739 2,165
41 Serveurperso/YuE2-GGUF — 7,537,625 8,847,449 23
42 ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF image-text-to-text 7,426,020 9,259,619 1,147
43 FacebookAI/roberta-large fill-mask 6,932,811 667,143,335 346
44 pyannote/speaker-diarization-3.1 automatic-speech-recognition 6,918,404 357,863,051 4,856
45 Qwen/Qwen3-VL-8B-InstructAlibaba image-text-to-text 6,915,010 78,692,156 1,237
46 ibm-granite/granite-embedding-small-english-r2 feature-extraction 6,776,901 33,413,479 113
47 intfloat/multilingual-e5-large feature-extraction 6,767,471 119,424,487 1,282
48 deepseek-ai/DeepSeek-V3.2DeepSeek text-generation 6,714,508 33,405,770 1,590
49 laion/clap-htsat-fused audio-classification 6,415,448 216,641,944 176
50 Qwen/Qwen3.5-9BAlibaba image-text-to-text 6,412,961 74,105,741 2,297
51 Qwen/Qwen3.5-4BAlibaba image-text-to-text 6,376,702 56,404,655 1,182
52 zai-org/GLM-5.3-FlashZ.ai image-text-to-text 6,313,299 12,573,351 3,029
53 FacebookAI/roberta-base fill-mask 6,299,171 662,792,559 700
54 MiniMaxAI/MiniMax-H3 image-text-to-video 6,140,874 15,690,081 6,530
55 amazon/chronos-bolt-small time-series-forecasting 6,127,478 46,395,160 67
56 distilbert/distilbert-base-uncased fill-mask 6,095,028 697,819,956 1,142
57 unsloth/Qwen3.8-27B-GGUF — 6,083,526 22,671,652 5,489
58 autogluon/chronos-bolt-small time-series-forecasting 6,019,165 162,682,784 86
59 openai/clip-vit-large-patch14OpenAI zero-shot-image-classification 6,006,155 1,201,492,076 2,136
60 microsoft/Florence-2-baseMicrosoft image-text-to-text 5,985,608 30,942,187 433
61 coqui/XTTS-v2 text-to-speech 5,965,681 132,667,274 3,931
62 google/gemma-4-26B-A4B-itGoogle image-text-to-text 5,944,935 74,528,422 1,731
63 openai/gpt-oss-20bOpenAI text-generation 5,939,725 107,425,867 5,263
64 Qwen/Qwen2.5-7B-InstructAlibaba text-generation 5,896,682 192,569,226 2,485
65 Qwen/Qwen2.5-VL-7B-InstructAlibaba image-text-to-text 5,886,476 108,837,296 1,763
66 pyannote/wespeaker-voxceleb-resnet34-LM — 5,815,981 405,446,734 218
67 openai/whisper-large-v3-turboOpenAI automatic-speech-recognition 5,664,908 126,560,599 3,499
68 ornith-ai/Ornith-1.5-9B-GGUF text-generation 5,435,094 13,789,963 625
69 microsoft/mdeberta-v3-baseMicrosoft fill-mask 5,357,805 60,875,164 261
70 Systran/faster-whisper-small automatic-speech-recognition 5,356,851 24,474,037 84
71 SamLowe/roberta-base-go_emotions text-classification 5,258,012 106,375,029 718
72 intfloat/multilingual-e5-base sentence-similarity 5,160,583 73,896,828 419
73 pyannote/segmentation-3.0 voice-activity-detection 5,119,817 406,510,051 2,496
74 pyannote/speaker-diarization-community-1 automatic-speech-recognition 5,102,899 37,551,741 3,737
75 Qwen/Qwen2.5-0.5B-InstructAlibaba text-generation 4,872,509 77,125,364 727
76 dphn/dolphin-2.9.1-yi-1.5-34b text-generation 4,771,746 84,774,201 90
77 Qwen/Qwen3.8-27B-FP8Alibaba image-text-to-text 4,731,404 16,036,405 955
78 Comfy-Org/Wan-2.1 — 4,667,254 64,232,389 1,038
79 Qwen/Qwen2.5-1.5B-InstructAlibaba text-generation 4,569,655 226,605,358 926
80 Qwen/Qwen3.6-35B-A3BAlibaba image-text-to-text 4,555,162 34,531,001 3,070
81 Xenova/all-MiniLM-L6-v2 feature-extraction 4,517,843 34,635,058 179
82 stabilityai/stable-diffusion-xl-base-1.0 text-to-image 4,493,100 130,329,554 8,427
83 facebook/opt-125mMeta text-generation 4,483,457 221,827,420 333
84 Comfy-Org/MOSS-Image — 4,467,577 4,467,577 582
85 meta-llama/Llama-3.1-8B-InstructMeta text-generation 4,463,539 192,376,330 8,760
86 laion/CLIP-ViT-B-32-laion2B-s34B-b79K zero-shot-image-classification 4,401,388 123,006,355 157
87 baidu/Unlimited-OCR image-text-to-text 4,241,405 12,765,657 4,939
88 Qwen/Qwen3-1.7BAlibaba text-generation 4,205,972 67,602,856 611
89 openai/whisper-smallOpenAI automatic-speech-recognition 4,166,276 140,883,700 645
90 Comfy-Org/Krea-2 text-to-image 4,143,236 18,000,690 786
91 google/gemma-4-E4B-itGoogle any-to-any 4,128,512 38,159,835 1,822
92 Qwen/Qwen3.6-35B-A3B-FP8Alibaba image-text-to-text 4,054,071 48,737,599 441
93 ornith-ai/Ornith-1.5-35B-A3B-GGUF text-generation 4,036,317 10,648,283 571
94 Qwen/Qwen3-VL-4B-InstructAlibaba image-text-to-text 3,936,514 32,849,793 523
95 Qwen/Qwen3.8-Flash-NextAlibaba image-text-to-text 3,929,164 5,789,760 6,728
96 facebook/esm2_t33_650M_UR50DMeta fill-mask 3,928,554 57,239,155 108
97 datasocietyco/bge-base-en-v1.5-course-recommender-v5 sentence-similarity 3,861,303 19,104,539 24
98 jonatasgrosman/wav2vec2-large-xlsr-53-russian automatic-speech-recognition 3,840,804 144,514,060 89
99 nvidia/Qwen3.6-35B-A3B-NVFP4 text-generation 3,788,274 40,934,747 673
100 ProsusAI/finbert text-classification 3,705,594 126,287,003 1,297

All-time counts that fell

shown, not smoothed

A fall shows from the second read. Users on the Hub’s forum report that an all-time count can fall from one day to the next. When one falls here, both values stay on this page.

What one download is

as the Hub counts it

The Hub counts a download each time a server sends one of a model’s count files, such as its config.json, to a GET or a HEAD request. A machine that loads a model a hundred times counts a hundred times, so the count measures use, not people. Each GGUF file counts on its own. The Hub’s own page gives the rules.

Sources

downloads.csvdownloads-labs.csv Every dataset → Status → Glossary →
  • Hugging Face: the Hub models API, read 3 Nov 2026 with no token, every model of each account above and the top 100 of the whole Hub. Each answer is kept byte for byte with its sha256. Each row of downloads.csv names its kept answer; the top model’s is f85c3f32d088.
  • Hugging Face: “Model download stats”, for how a download is counted.