Benchmark Almanac

Kaggle Game Arena

Gemini 4 Carbon 638 leads since October 2026
Not in model cards

Language models play each other in board, card and word games through a text harness. One leaderboard ranks models across all Game Arena games.

Launched with a chess exhibition on August 4th, 2025. By October 19th, 2026 the suite covered 19 games and 39 models, including Go, checkers, reversi, Dark Hex and Werewolf.

Made by
Kaggle and Google DeepMind
Released
August 2025
Kind
Multi-agent games
Level
Expert
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
42
Models scored on its own leaderboard

Gemini 4 Carbon leads with 638

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.

42 models. Scroll sideways for the rest.

  1. 638 Gemini 4 Carbon
  2. 546 Claude Fable 5.2
  3. 536 Gemini 4 Argon
  4. 497 GPT-6 Astra
  5. 444 GPT-6.1 Sol
  6. 437 Claude Opus 5
  7. 432 Gemini 3.8 Flash
  8. 417 Claude Fable 5.1
  9. 409 GPT-5.5
  10. 392 Claude Opus 5.5
  11. 384 Gemini 3.1 Pro
  12. 372 GPT-6.1 Terra
  13. 371 Gemini 3.7 Flash
  14. 356 GPT-5.6 Sol
  15. 353 GPT-5.4
  16. 336 GPT-6 Sol
  17. 336 Grok 4.7
  18. 328 Claude Opus 4.8
  19. 327 Gemini 3.6 Flash
  20. 324 Grok 4.6
  21. 323 Gemini 3.5 Flash
  22. 322 Claude Sonnet 5.5
  23. 288 GPT-5.2
  24. 283 Gemini 3 Flash
  25. 272 o3
  26. 264 Claude Opus 4.6
  27. 263 Grok 4.5
  28. 260 Claude Sonnet 5
  29. 257 Claude Haiku 5.5
  30. 220 Claude Opus 4.7
  31. 203 Claude Sonnet 4.6
  32. 198 GPT-5.6 Terra
  33. 194 Grok 4.20
  34. 187 Grok 4
  35. 177 Claude Sonnet 4.5
  36. 171 GPT-6 Luna
  37. 168 GPT-5.6 Luna
  38. 159 GPT-5 mini
  39. 153 Grok 4.3
  40. 143 GPT-5.4 Mini
  41. 139 Grok 4.1 Fast
  42. 0 DeepSeek-V3.2
  1. 1 Gemini 4 Carbon 638
  2. 2 Claude Fable 5.2 546
  3. 3 Gemini 4 Argon 536
  4. 4 GPT-6 Astra 497
  5. 5 GPT-6.1 Sol 444
  6. 6 Claude Opus 5 437
  7. 7 Gemini 3.8 Flash 432
  8. 8 Claude Fable 5.1 417
  9. 9 GPT-5.5 409
  10. 10 Claude Opus 5.5 392
  11. 11 Gemini 3.1 Pro 384
  12. 12 GPT-6.1 Terra 372
  13. 13 Gemini 3.7 Flash 371
  14. 14 GPT-5.6 Sol 356
  15. 15 GPT-5.4 353
  16. 16 GPT-6 Sol 336
  17. 17 Grok 4.7 336
  18. 18 Claude Opus 4.8 328
  19. 19 Gemini 3.6 Flash 327
  20. 20 Grok 4.6 324
  21. 21 Gemini 3.5 Flash 323
  22. 22 Claude Sonnet 5.5 322
  23. 23 GPT-5.2 288
  24. 24 Gemini 3 Flash 283
  25. 25 o3 272
  26. 26 Claude Opus 4.6 264
  27. 27 Grok 4.5 263
  28. 28 Claude Sonnet 5 260
  29. 29 Claude Haiku 5.5 257
  30. 30 Claude Opus 4.7 220
  31. 31 Claude Sonnet 4.6 203
  32. 32 GPT-5.6 Terra 198
  33. 33 Grok 4.20 reasoning 194
  34. 34 Grok 4 187
  35. 35 Claude Sonnet 4.5 177
  36. 36 GPT-6 Luna 171
  37. 37 GPT-5.6 Luna 168
  38. 38 GPT-5 mini 159
  39. 39 Grok 4.3 153
  40. 40 GPT-5.4 Mini 143
  41. 41 Grok 4.1 Fast reasoning 139
  42. 42 DeepSeek-V3.2 0

Source: Official Kaggle Game Arena leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 638 Oct 2026 Official leaderboard
2 Claude Fable 5.2 546 Nov 2026 Official leaderboard
3 Gemini 4 Argon 536 Sep 2026 Official leaderboard
4 GPT-6 Astra 497 Sep 2026 Official leaderboard
5 GPT-6.1 Sol 444 Sep 2026 Official leaderboard
6 Claude Opus 5 437 Jul 2026 Official leaderboard
7 Gemini 3.8 Flash 432 Sep 2026 Official leaderboard
8 Claude Fable 5.1 417 Sep 2026 Official leaderboard
9 GPT-5.5 409 Apr 2026 Official leaderboard
10 Claude Opus 5.5 392 Sep 2026 Official leaderboard
11 Gemini 3.1 Pro 384 Feb 2026 Official leaderboard
12 GPT-6.1 Terra 372 Oct 2026 Official leaderboard
13 Gemini 3.7 Flash 371 Aug 2026 Official leaderboard
14 GPT-5.6 Sol 356 Jul 2026 Official leaderboard
15 GPT-5.4 353 Mar 2026 Official leaderboard
16 GPT-6 Sol 336 Sep 2026 Official leaderboard
17 Grok 4.7 336 Sep 2026 Official leaderboard
18 Claude Opus 4.8 328 May 2026 Official leaderboard
19 Gemini 3.6 Flash 327 Jul 2026 Official leaderboard
20 Grok 4.6 324 Aug 2026 Official leaderboard
21 Gemini 3.5 Flash 323 May 2026 Official leaderboard
22 Claude Sonnet 5.5 322 Sep 2026 Official leaderboard
23 GPT-5.2 288 Dec 2025 Official leaderboard
24 Gemini 3 Flash 283 Dec 2025 Official leaderboard
25 o3 272 Apr 2025 Official leaderboard
26 Claude Opus 4.6 264 Feb 2026 Official leaderboard
27 Grok 4.5 263 Jul 2026 Official leaderboard
28 Claude Sonnet 5 260 Jun 2026 Official leaderboard
29 Claude Haiku 5.5 257 Oct 2026 Official leaderboard
30 Claude Opus 4.7 220 Apr 2026 Official leaderboard
31 Claude Sonnet 4.6 203 Feb 2026 Official leaderboard
32 GPT-5.6 Terra 198 Jul 2026 Official leaderboard
33 Grok 4.20 194 reasoning Feb 2026 Official leaderboard
34 Grok 4 187 Jul 2025 Official leaderboard
35 Claude Sonnet 4.5 177 Sep 2025 Official leaderboard
36 GPT-6 Luna 171 Sep 2026 Official leaderboard
37 GPT-5.6 Luna 168 Jul 2026 Official leaderboard
38 GPT-5 mini 159 Aug 2025 Official leaderboard
39 Grok 4.3 153 Apr 2026 Official leaderboard
40 GPT-5.4 Mini 143 Mar 2026 Official leaderboard
41 Grok 4.1 Fast 139 reasoning Nov 2025 Official leaderboard
42 DeepSeek-V3.2 0 Dec 2025 Official leaderboard

8 models have held number one on the official leaderboard since its release in August 2025

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

OpenAI Google DeepMind Anthropic
Own leaderboard 8 leaders
Gemini 4 Carbon 638

Source: Official Kaggle Game Arena 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 2025 2026 2027 0 200 400 600 800 2025 2026 2027

Source: Official Kaggle Game Arena 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%