Benchmark Almanac

Game Arena Chess

Gemini 3.8 Flash 1,453 leads since Sep 2026 Not in model cards

Language models play full chess games against each other with the board given as FEN and PGN text. Ratings come from many games between each pair.

The 2026-02-02 leaderboard is labeled v2 of the chess text task. Kaggle also runs Chess Openings (11 models) and a Chess Suite (19 models). 39 models on 2026-10-19.

Made by Kaggle and Google DeepMind Released Aug 2025 Puzzles and chess All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Ranked by matches Models play against each other, and the results become ratings.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
google-deepmind/game_arena 117 stars Leaderboard kaggle.com
39
Models scored on its own leaderboard

Gemini 3.8 Flash leads with 1,453

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

39 models. Scroll sideways for the rest.

  1. 1,453 Gemini 3.8 Flash
  2. 1,427 GPT-6 Astra
  3. 1,398 Gemini 3.7 Flash
  4. 1,381 Gemini 4 Argon
  5. 1,343 Gemini 3.5 Flash
  6. 1,324 Gemini 3.1 Pro
  7. 1,257 GPT-5.5
  8. 1,238 Gemini 3.6 Flash
  9. 1,230 GPT-6.1 Sol
  10. 1,212 Gemini 3 Flash
  11. 1,134 GPT-5.6 Sol
  12. 1,007 GPT-6 Sol
  13. 962 Claude Opus 5.5
  14. 923 o3
  15. 896 GPT-5.4
  16. 844 Grok 4.7
  17. 825 GPT-5.2
  18. 808 Grok 4.6
  19. 781 Claude Sonnet 5.5
  20. 760 Claude Fable 5.1
  21. 755 Grok 4
  22. 725 GPT-5.6 Terra
  23. 700 GPT-5.6 Luna
  24. 690 Claude Opus 5
  25. 665 Grok 4.20
  26. 643 Grok 4.5
  27. 622 Claude Opus 4.8
  28. 606 Grok 4.1 Fast
  29. 594 Claude Haiku 5.5
  30. 591 Grok 4.3
  31. 585 Claude Sonnet 5
  32. 537 GPT-5 mini
  33. 449 GPT-6 Luna
  34. 435 GPT-5.4 Mini
  35. 284 Claude Opus 4.7
  36. 98 Claude Sonnet 4.6
  37. 84 DeepSeek-V3.2
  38. 50 Claude Opus 4.6
  39. 0 Claude Sonnet 4.5
  1. 1 Gemini 3.8 Flash 1,453
  2. 2 GPT-6 Astra 1,427
  3. 3 Gemini 3.7 Flash 1,398
  4. 4 Gemini 4 Argon 1,381
  5. 5 Gemini 3.5 Flash 1,343
  6. 6 Gemini 3.1 Pro 1,324
  7. 7 GPT-5.5 1,257
  8. 8 Gemini 3.6 Flash 1,238
  9. 9 GPT-6.1 Sol 1,230
  10. 10 Gemini 3 Flash 1,212
  11. 11 GPT-5.6 Sol 1,134
  12. 12 GPT-6 Sol 1,007
  13. 13 Claude Opus 5.5 962
  14. 14 o3 923
  15. 15 GPT-5.4 896
  16. 16 Grok 4.7 844
  17. 17 GPT-5.2 825
  18. 18 Grok 4.6 808
  19. 19 Claude Sonnet 5.5 781
  20. 20 Claude Fable 5.1 760
  21. 21 Grok 4 755
  22. 22 GPT-5.6 Terra 725
  23. 23 GPT-5.6 Luna 700
  24. 24 Claude Opus 5 690
  25. 25 Grok 4.20 reasoning 665
  26. 26 Grok 4.5 643
  27. 27 Claude Opus 4.8 622
  28. 28 Grok 4.1 Fast reasoning 606
  29. 29 Claude Haiku 5.5 594
  30. 30 Grok 4.3 591
  31. 31 Claude Sonnet 5 585
  32. 32 GPT-5 mini 537
  33. 33 GPT-6 Luna 449
  34. 34 GPT-5.4 Mini 435
  35. 35 Claude Opus 4.7 284
  36. 36 Claude Sonnet 4.6 98
  37. 37 DeepSeek-V3.2 84
  38. 38 Claude Opus 4.6 50
  39. 39 Claude Sonnet 4.5 0
#ModelScoreSettingDateSource
1 Gemini 3.8 Flash 1,453 Sep 2026 Official leaderboard
2 GPT-6 Astra 1,427 Sep 2026 Official leaderboard
3 Gemini 3.7 Flash 1,398 Aug 2026 Official leaderboard
4 Gemini 4 Argon 1,381 Sep 2026 Official leaderboard
5 Gemini 3.5 Flash 1,343 May 2026 Official leaderboard
6 Gemini 3.1 Pro 1,324 Feb 2026 Official leaderboard
7 GPT-5.5 1,257 Apr 2026 Official leaderboard
8 Gemini 3.6 Flash 1,238 Jul 2026 Official leaderboard
9 GPT-6.1 Sol 1,230 Sep 2026 Official leaderboard
10 Gemini 3 Flash 1,212 Dec 2025 Official leaderboard
11 GPT-5.6 Sol 1,134 Jul 2026 Official leaderboard
12 GPT-6 Sol 1,007 Sep 2026 Official leaderboard
13 Claude Opus 5.5 962 Sep 2026 Official leaderboard
14 o3 923 Apr 2025 Official leaderboard
15 GPT-5.4 896 Mar 2026 Official leaderboard
16 Grok 4.7 844 Sep 2026 Official leaderboard
17 GPT-5.2 825 Dec 2025 Official leaderboard
18 Grok 4.6 808 Aug 2026 Official leaderboard
19 Claude Sonnet 5.5 781 Sep 2026 Official leaderboard
20 Claude Fable 5.1 760 Sep 2026 Official leaderboard
21 Grok 4 755 Jul 2025 Official leaderboard
22 GPT-5.6 Terra 725 Jul 2026 Official leaderboard
23 GPT-5.6 Luna 700 Jul 2026 Official leaderboard
24 Claude Opus 5 690 Jul 2026 Official leaderboard
25 Grok 4.20 665 reasoning Feb 2026 Official leaderboard
26 Grok 4.5 643 Jul 2026 Official leaderboard
27 Claude Opus 4.8 622 May 2026 Official leaderboard
28 Grok 4.1 Fast 606 reasoning Nov 2025 Official leaderboard
29 Claude Haiku 5.5 594 Oct 2026 Official leaderboard
30 Grok 4.3 591 Apr 2026 Official leaderboard
31 Claude Sonnet 5 585 Jun 2026 Official leaderboard
32 GPT-5 mini 537 Aug 2025 Official leaderboard
33 GPT-6 Luna 449 Sep 2026 Official leaderboard
34 GPT-5.4 Mini 435 Mar 2026 Official leaderboard
35 Claude Opus 4.7 284 Apr 2026 Official leaderboard
36 Claude Sonnet 4.6 98 Feb 2026 Official leaderboard
37 DeepSeek-V3.2 84 Dec 2025 Official leaderboard
38 Claude Opus 4.6 50 Feb 2026 Official leaderboard
39 Claude Sonnet 4.5 0 Sep 2025 Official leaderboard

6 models have held number one since its release in Aug 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI Google DeepMind
Own leaderboard 6 leaders
Gemini 3.8 Flash 1,453

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
0 500 1,000 1,500 2025 2026 0 500 1,000 1,500 2025 2026

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

More in Games and long tasks

All 68 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6 Astra 100%
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Scores come from in-game progress or outcomes per title. Nemotron 3 Nano 4B 22.9
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. Gemini 3 Flash 41.9%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. Measures game progress and harmful actions per 100 steps. Muse Spark 2.40