Benchmark Almanac

Game Arena Clobber

Gemini 4 Carbon 819 leads since October 2026
Not in model cards

Language models play Clobber, where every move captures an adjacent enemy stone. The player who makes the last legal move wins.

Kaggle published the leaderboard on June 25th, 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
June 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 819

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. 819 Gemini 4 Carbon
  2. 757 GPT-6 Astra
  3. 695 Gemini 4 Argon
  4. 653 GPT-6.1 Sol
  5. 640 Claude Fable 5.2
  6. 560 Claude Opus 5
  7. 519 Claude Fable 5.1
  8. 489 Claude Opus 5.5
  9. 483 Gemini 3.1 Pro
  10. 477 Gemini 3.8 Flash
  11. 449 GPT-5.4
  12. 414 GPT-5.5
  13. 411 GPT-6.1 Terra
  14. 390 Gemini 3.7 Flash
  15. 348 GPT-5.2
  16. 339 GPT-5.6 Sol
  17. 330 Grok 4.7
  18. 320 Claude Opus 4.6
  19. 316 GPT-6 Sol
  20. 304 Claude Opus 4.8
  21. 290 Gemini 3.6 Flash
  22. 262 Claude Sonnet 4.6
  23. 247 Claude Sonnet 5
  24. 226 Claude Sonnet 4.5
  25. 211 Gemini 3.5 Flash
  26. 170 Claude Sonnet 5.5
  27. 167 Grok 4.6
  28. 160 GPT-5.6 Terra
  29. 153 Grok 4.5
  30. 150 GPT-5.4 Mini
  31. 144 o3
  32. 133 Claude Haiku 5.5
  33. 110 Grok 4.20
  34. 66 GPT-6 Luna
  35. 43 GPT-5 mini
  36. 37 GPT-5.6 Luna
  37. 31 Claude Opus 4.7
  38. 22 Gemini 3 Flash
  39. 0 Grok 4.3
  1. 1 Gemini 4 Carbon 819
  2. 2 GPT-6 Astra 757
  3. 3 Gemini 4 Argon 695
  4. 4 GPT-6.1 Sol 653
  5. 5 Claude Fable 5.2 640
  6. 6 Claude Opus 5 560
  7. 7 Claude Fable 5.1 519
  8. 8 Claude Opus 5.5 489
  9. 9 Gemini 3.1 Pro 483
  10. 10 Gemini 3.8 Flash 477
  11. 11 GPT-5.4 449
  12. 12 GPT-5.5 414
  13. 13 GPT-6.1 Terra 411
  14. 14 Gemini 3.7 Flash 390
  15. 15 GPT-5.2 348
  16. 16 GPT-5.6 Sol 339
  17. 17 Grok 4.7 330
  18. 18 Claude Opus 4.6 320
  19. 19 GPT-6 Sol 316
  20. 20 Claude Opus 4.8 304
  21. 21 Gemini 3.6 Flash 290
  22. 22 Claude Sonnet 4.6 262
  23. 23 Claude Sonnet 5 247
  24. 24 Claude Sonnet 4.5 226
  25. 25 Gemini 3.5 Flash 211
  26. 26 Claude Sonnet 5.5 170
  27. 27 Grok 4.6 167
  28. 28 GPT-5.6 Terra 160
  29. 29 Grok 4.5 153
  30. 30 GPT-5.4 Mini 150
  31. 31 o3 144
  32. 32 Claude Haiku 5.5 133
  33. 33 Grok 4.20 reasoning 110
  34. 34 GPT-6 Luna 66
  35. 35 GPT-5 mini 43
  36. 36 GPT-5.6 Luna 37
  37. 37 Claude Opus 4.7 31
  38. 38 Gemini 3 Flash 22
  39. 39 Grok 4.3 0

Source: Official Game Arena Clobber leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 819 Oct 2026 Official leaderboard
2 GPT-6 Astra 757 Sep 2026 Official leaderboard
3 Gemini 4 Argon 695 Sep 2026 Official leaderboard
4 GPT-6.1 Sol 653 Sep 2026 Official leaderboard
5 Claude Fable 5.2 640 Nov 2026 Official leaderboard
6 Claude Opus 5 560 Jul 2026 Official leaderboard
7 Claude Fable 5.1 519 Sep 2026 Official leaderboard
8 Claude Opus 5.5 489 Sep 2026 Official leaderboard
9 Gemini 3.1 Pro 483 Feb 2026 Official leaderboard
10 Gemini 3.8 Flash 477 Sep 2026 Official leaderboard
11 GPT-5.4 449 Mar 2026 Official leaderboard
12 GPT-5.5 414 Apr 2026 Official leaderboard
13 GPT-6.1 Terra 411 Oct 2026 Official leaderboard
14 Gemini 3.7 Flash 390 Aug 2026 Official leaderboard
15 GPT-5.2 348 Dec 2025 Official leaderboard
16 GPT-5.6 Sol 339 Jul 2026 Official leaderboard
17 Grok 4.7 330 Sep 2026 Official leaderboard
18 Claude Opus 4.6 320 Feb 2026 Official leaderboard
19 GPT-6 Sol 316 Sep 2026 Official leaderboard
20 Claude Opus 4.8 304 May 2026 Official leaderboard
21 Gemini 3.6 Flash 290 Jul 2026 Official leaderboard
22 Claude Sonnet 4.6 262 Feb 2026 Official leaderboard
23 Claude Sonnet 5 247 Jun 2026 Official leaderboard
24 Claude Sonnet 4.5 226 Sep 2025 Official leaderboard
25 Gemini 3.5 Flash 211 May 2026 Official leaderboard
26 Claude Sonnet 5.5 170 Sep 2026 Official leaderboard
27 Grok 4.6 167 Aug 2026 Official leaderboard
28 GPT-5.6 Terra 160 Jul 2026 Official leaderboard
29 Grok 4.5 153 Jul 2026 Official leaderboard
30 GPT-5.4 Mini 150 Mar 2026 Official leaderboard
31 o3 144 Apr 2025 Official leaderboard
32 Claude Haiku 5.5 133 Oct 2026 Official leaderboard
33 Grok 4.20 110 reasoning Feb 2026 Official leaderboard
34 GPT-6 Luna 66 Sep 2026 Official leaderboard
35 GPT-5 mini 43 Aug 2025 Official leaderboard
36 GPT-5.6 Luna 37 Jul 2026 Official leaderboard
37 Claude Opus 4.7 31 Apr 2026 Official leaderboard
38 Gemini 3 Flash 22 Dec 2025 Official leaderboard
39 Grok 4.3 0 Apr 2026 Official leaderboard

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

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

Google DeepMind Anthropic OpenAI
Own leaderboard 4 leaders
Gemini 4 Carbon 819

Source: Official Game Arena Clobber 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
0 200 400 600 800 1,000 2025 2026 2027 0 200 400 600 800 1,000 2025 2026 2027

Source: Official Game Arena Clobber 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%