Benchmark Almanac

Game Arena Reversi

Gemini 4 Carbon 1,306 leads since October 2026
Not in model cards

Language models play Reversi, also called Othello, flipping the opponent's disks that they flank in a line.

Kaggle published the leaderboard on August 11th, 2026 with 36 models on November 1st, 2026. Gameplay logs are a public Kaggle dataset. Part of the Kaggle Game Arena suite.

Made by
Kaggle and Google DeepMind
Released
August 2026
Kind
Puzzles and chess
Level
Everyday
Scored
Ranked by matches
Answers
Agent task
Tasks
All public
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
39
Models scored on its own leaderboard

Gemini 4 Carbon leads with 1,306

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

39 models. Scroll sideways for the rest.

  1. 1,306 Gemini 4 Carbon
  2. 1,121 Gemini 4 Argon
  3. 935 Gemini 3.8 Flash
  4. 920 GPT-6 Astra
  5. 882 Claude Fable 5.2
  6. 871 Claude Opus 5
  7. 818 Gemini 3.7 Flash
  8. 796 GPT-6.1 Sol
  9. 795 GPT-5.5
  10. 756 GPT-5.4
  11. 743 Claude Fable 5.1
  12. 714 Gemini 3.5 Flash
  13. 710 Gemini 3.6 Flash
  14. 710 Gemini 3.1 Pro
  15. 694 Claude Opus 5.5
  16. 690 Claude Haiku 5.5
  17. 625 GPT-5.2
  18. 585 GPT-5.6 Sol
  19. 552 GPT-6.1 Terra
  20. 540 Claude Sonnet 5.5
  21. 494 GPT-6 Sol
  22. 454 o3
  23. 453 Grok 4.7
  24. 375 Claude Opus 4.8
  25. 367 Grok 4.6
  26. 330 Gemini 3 Flash
  27. 324 Claude Opus 4.6
  28. 253 Claude Sonnet 5
  29. 240 Grok 4.5
  30. 215 Claude Sonnet 4.6
  31. 202 GPT-5 mini
  32. 181 GPT-6 Luna
  33. 150 Claude Sonnet 4.5
  34. 148 GPT-5.6 Terra
  35. 122 Grok 4.20
  36. 91 GPT-5.4 Mini
  37. 42 Claude Opus 4.7
  38. 14 Grok 4.3
  39. 0 GPT-5.6 Luna
  1. 1 Gemini 4 Carbon 1,306
  2. 2 Gemini 4 Argon 1,121
  3. 3 Gemini 3.8 Flash 935
  4. 4 GPT-6 Astra 920
  5. 5 Claude Fable 5.2 882
  6. 6 Claude Opus 5 871
  7. 7 Gemini 3.7 Flash 818
  8. 8 GPT-6.1 Sol 796
  9. 9 GPT-5.5 795
  10. 10 GPT-5.4 756
  11. 11 Claude Fable 5.1 743
  12. 12 Gemini 3.5 Flash 714
  13. 13 Gemini 3.6 Flash 710
  14. 14 Gemini 3.1 Pro 710
  15. 15 Claude Opus 5.5 694
  16. 16 Claude Haiku 5.5 690
  17. 17 GPT-5.2 625
  18. 18 GPT-5.6 Sol 585
  19. 19 GPT-6.1 Terra 552
  20. 20 Claude Sonnet 5.5 540
  21. 21 GPT-6 Sol 494
  22. 22 o3 454
  23. 23 Grok 4.7 453
  24. 24 Claude Opus 4.8 375
  25. 25 Grok 4.6 367
  26. 26 Gemini 3 Flash 330
  27. 27 Claude Opus 4.6 324
  28. 28 Claude Sonnet 5 253
  29. 29 Grok 4.5 240
  30. 30 Claude Sonnet 4.6 215
  31. 31 GPT-5 mini 202
  32. 32 GPT-6 Luna 181
  33. 33 Claude Sonnet 4.5 150
  34. 34 GPT-5.6 Terra 148
  35. 35 Grok 4.20 reasoning 122
  36. 36 GPT-5.4 Mini 91
  37. 37 Claude Opus 4.7 42
  38. 38 Grok 4.3 14
  39. 39 GPT-5.6 Luna 0

Source: Official Game Arena Reversi leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 1,306 Oct 2026 Official leaderboard
2 Gemini 4 Argon 1,121 Sep 2026 Official leaderboard
3 Gemini 3.8 Flash 935 Sep 2026 Official leaderboard
4 GPT-6 Astra 920 Sep 2026 Official leaderboard
5 Claude Fable 5.2 882 Nov 2026 Official leaderboard
6 Claude Opus 5 871 Jul 2026 Official leaderboard
7 Gemini 3.7 Flash 818 Aug 2026 Official leaderboard
8 GPT-6.1 Sol 796 Sep 2026 Official leaderboard
9 GPT-5.5 795 Apr 2026 Official leaderboard
10 GPT-5.4 756 Mar 2026 Official leaderboard
11 Claude Fable 5.1 743 Sep 2026 Official leaderboard
12 Gemini 3.5 Flash 714 May 2026 Official leaderboard
13 Gemini 3.6 Flash 710 Jul 2026 Official leaderboard
14 Gemini 3.1 Pro 710 Feb 2026 Official leaderboard
15 Claude Opus 5.5 694 Sep 2026 Official leaderboard
16 Claude Haiku 5.5 690 Oct 2026 Official leaderboard
17 GPT-5.2 625 Dec 2025 Official leaderboard
18 GPT-5.6 Sol 585 Jul 2026 Official leaderboard
19 GPT-6.1 Terra 552 Oct 2026 Official leaderboard
20 Claude Sonnet 5.5 540 Sep 2026 Official leaderboard
21 GPT-6 Sol 494 Sep 2026 Official leaderboard
22 o3 454 Apr 2025 Official leaderboard
23 Grok 4.7 453 Sep 2026 Official leaderboard
24 Claude Opus 4.8 375 May 2026 Official leaderboard
25 Grok 4.6 367 Aug 2026 Official leaderboard
26 Gemini 3 Flash 330 Dec 2025 Official leaderboard
27 Claude Opus 4.6 324 Feb 2026 Official leaderboard
28 Claude Sonnet 5 253 Jun 2026 Official leaderboard
29 Grok 4.5 240 Jul 2026 Official leaderboard
30 Claude Sonnet 4.6 215 Feb 2026 Official leaderboard
31 GPT-5 mini 202 Aug 2025 Official leaderboard
32 GPT-6 Luna 181 Sep 2026 Official leaderboard
33 Claude Sonnet 4.5 150 Sep 2025 Official leaderboard
34 GPT-5.6 Terra 148 Jul 2026 Official leaderboard
35 Grok 4.20 122 reasoning Feb 2026 Official leaderboard
36 GPT-5.4 Mini 91 Mar 2026 Official leaderboard
37 Claude Opus 4.7 42 Apr 2026 Official leaderboard
38 Grok 4.3 14 Apr 2026 Official leaderboard
39 GPT-5.6 Luna 0 Jul 2026 Official leaderboard

4 models have held number one on the official leaderboard since its release in August 2026

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

Anthropic Google DeepMind
Own leaderboard 4 leaders
Gemini 4 Carbon 1,306

Source: Official Game Arena Reversi leaderboard

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 2027 0 500 1,000 1,500 2025 2026 2027

Source: Official Game Arena Reversi leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Games and long tasks

All 82 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 99.95%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • 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%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • 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%