Anthropic · proprietary on Arena · first on Arena 6 February 2026
claude-opus-4-6-thinking
197 days in first place
It held first place on Arena in Text (143 days), WebDev (35 days) and Vision (19 days).
Text: 143 days in first place
last on the board 11 August 2026 1st10th
- On the board
- 6 February 2026 to 11 August 2026
- Score
- 1503 (1499–1506) on 11 August 2026 · 72,425 votes
- First place
-
- 6 February 2026 to 19 February 2026 · 13 days
- 18 March 2026 to 26 July 2026 · 130 days
- Catch-up
- It left the board on 11 August 2026, 186 days after it first took first place. No open-weights model had matched it.
WebDev: 35 days in first place
last on the board 11 August 2026 1st10th
- On the board
- 9 February 2026 to 11 August 2026
- Score
- 1546 (1540–1552) on 11 August 2026 · 16,054 votes
- First place
-
- 9 February 2026 to 17 February 2026 · 8 days
- 4 March 2026 to 16 March 2026 · 12 days
- 2 April 2026 to 17 April 2026 · 15 days
- Catch-up
- Open weights matched it on 16 June 2026, 127 days after it first took first place: glm-5.2 (max).
Vision: 19 days in first place
last on the board 6 August 2026 1st10th
- On the board
- 26 March 2026 to 6 August 2026
- Score
- 1316 (1309–1323) on 6 August 2026 · 20,310 votes
- First place
-
- 10 April 2026 to 19 April 2026 · 9 days
- 27 April 2026 to 7 May 2026 · 10 days
- Catch-up
- It left the board on 6 August 2026, 118 days after it first took first place. No open-weights model had matched it.
Epoch AI: Claude Opus 4.6
released 5 February 2026 · API access- Chess Puzzles 21.0%
- EBR-bench 12.7%
- FrontierMath-2025-02-28-Private 40.7%
- FrontierMath-Tier-4-2025-07-01-Private 22.9%
- FrontierMath-Tier-4-v2-Private 26.8%
- FrontierMath-Tiers-1-3-v2-Private 66.0%
- Furniture Assembly 28.3%
- GPQA diamond 90.5%
- Mystery Game Puzzles 25.0%
- OTIS Mock AIME 2024-2025 94.4%
- SWE-Bench verified 78.7%
- SimpleQA Verified 47.0%
When each vendor retires it
every model’s dates- Anthropic not before 5 Feb 2027 claude-opus-4-6
- Azure AI Foundry 2 Feb 2027 claude-opus-4-6 (version 1)
Sources
Arena scores come from people voting between two anonymous answers. First place is the highest score; the top two are often within each other’s margin of error. LMArena, Arena leaderboard dataset, revision 4491b06, retrieved 30 October 2026; Epoch AI, Capabilities & benchmarking, retrieved 30 October 2026. Both CC BY 4.0. Retirement dates: each vendor’s own deprecation page, quoted and kept, listed with the record.