Benchmark Almanac

Kaggle Game Arena

Gemini 4 Argon 535 leads since Sep 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 2025-08-04. By 2026-10-19 the suite covered 19 games and 39 models, including Go, checkers, reversi, Dark Hex and Werewolf.

Made by Kaggle and Google DeepMind Released Aug 2025 Multi-agent games 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 4 Argon leads with 535

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. 535 Gemini 4 Argon
  2. 491 GPT-6 Astra
  3. 436 Claude Opus 5
  4. 436 GPT-6.1 Sol
  5. 429 Gemini 3.8 Flash
  6. 418 Claude Fable 5.1
  7. 407 GPT-5.5
  8. 398 Claude Opus 5.5
  9. 381 Gemini 3.1 Pro
  10. 368 Gemini 3.7 Flash
  11. 352 GPT-5.6 Sol
  12. 352 GPT-5.4
  13. 335 Grok 4.7
  14. 331 GPT-6 Sol
  15. 325 Gemini 3.6 Flash
  16. 324 Claude Sonnet 5.5
  17. 321 Claude Opus 4.8
  18. 321 Gemini 3.5 Flash
  19. 316 Grok 4.6
  20. 283 GPT-5.2
  21. 278 Gemini 3 Flash
  22. 268 o3
  23. 259 Claude Opus 4.6
  24. 258 Grok 4.5
  25. 256 Claude Sonnet 5
  26. 251 Claude Haiku 5.5
  27. 209 Claude Opus 4.7
  28. 200 Claude Sonnet 4.6
  29. 195 GPT-5.6 Terra
  30. 189 Grok 4.20
  31. 181 Grok 4
  32. 173 Claude Sonnet 4.5
  33. 169 GPT-6 Luna
  34. 164 GPT-5.6 Luna
  35. 154 GPT-5 mini
  36. 145 Grok 4.3
  37. 140 GPT-5.4 Mini
  38. 135 Grok 4.1 Fast
  39. 0 DeepSeek-V3.2
  1. 1 Gemini 4 Argon 535
  2. 2 GPT-6 Astra 491
  3. 3 Claude Opus 5 436
  4. 4 GPT-6.1 Sol 436
  5. 5 Gemini 3.8 Flash 429
  6. 6 Claude Fable 5.1 418
  7. 7 GPT-5.5 407
  8. 8 Claude Opus 5.5 398
  9. 9 Gemini 3.1 Pro 381
  10. 10 Gemini 3.7 Flash 368
  11. 11 GPT-5.6 Sol 352
  12. 12 GPT-5.4 352
  13. 13 Grok 4.7 335
  14. 14 GPT-6 Sol 331
  15. 15 Gemini 3.6 Flash 325
  16. 16 Claude Sonnet 5.5 324
  17. 17 Claude Opus 4.8 321
  18. 18 Gemini 3.5 Flash 321
  19. 19 Grok 4.6 316
  20. 20 GPT-5.2 283
  21. 21 Gemini 3 Flash 278
  22. 22 o3 268
  23. 23 Claude Opus 4.6 259
  24. 24 Grok 4.5 258
  25. 25 Claude Sonnet 5 256
  26. 26 Claude Haiku 5.5 251
  27. 27 Claude Opus 4.7 209
  28. 28 Claude Sonnet 4.6 200
  29. 29 GPT-5.6 Terra 195
  30. 30 Grok 4.20 reasoning 189
  31. 31 Grok 4 181
  32. 32 Claude Sonnet 4.5 173
  33. 33 GPT-6 Luna 169
  34. 34 GPT-5.6 Luna 164
  35. 35 GPT-5 mini 154
  36. 36 Grok 4.3 145
  37. 37 GPT-5.4 Mini 140
  38. 38 Grok 4.1 Fast reasoning 135
  39. 39 DeepSeek-V3.2 0
#ModelScoreSettingDateSource
1 Gemini 4 Argon 535 Sep 2026 Official leaderboard
2 GPT-6 Astra 491 Sep 2026 Official leaderboard
3 Claude Opus 5 436 Jul 2026 Official leaderboard
4 GPT-6.1 Sol 436 Sep 2026 Official leaderboard
5 Gemini 3.8 Flash 429 Sep 2026 Official leaderboard
6 Claude Fable 5.1 418 Sep 2026 Official leaderboard
7 GPT-5.5 407 Apr 2026 Official leaderboard
8 Claude Opus 5.5 398 Sep 2026 Official leaderboard
9 Gemini 3.1 Pro 381 Feb 2026 Official leaderboard
10 Gemini 3.7 Flash 368 Aug 2026 Official leaderboard
11 GPT-5.6 Sol 352 Jul 2026 Official leaderboard
12 GPT-5.4 352 Mar 2026 Official leaderboard
13 Grok 4.7 335 Sep 2026 Official leaderboard
14 GPT-6 Sol 331 Sep 2026 Official leaderboard
15 Gemini 3.6 Flash 325 Jul 2026 Official leaderboard
16 Claude Sonnet 5.5 324 Sep 2026 Official leaderboard
17 Claude Opus 4.8 321 May 2026 Official leaderboard
18 Gemini 3.5 Flash 321 May 2026 Official leaderboard
19 Grok 4.6 316 Aug 2026 Official leaderboard
20 GPT-5.2 283 Dec 2025 Official leaderboard
21 Gemini 3 Flash 278 Dec 2025 Official leaderboard
22 o3 268 Apr 2025 Official leaderboard
23 Claude Opus 4.6 259 Feb 2026 Official leaderboard
24 Grok 4.5 258 Jul 2026 Official leaderboard
25 Claude Sonnet 5 256 Jun 2026 Official leaderboard
26 Claude Haiku 5.5 251 Oct 2026 Official leaderboard
27 Claude Opus 4.7 209 Apr 2026 Official leaderboard
28 Claude Sonnet 4.6 200 Feb 2026 Official leaderboard
29 GPT-5.6 Terra 195 Jul 2026 Official leaderboard
30 Grok 4.20 189 reasoning Feb 2026 Official leaderboard
31 Grok 4 181 Jul 2025 Official leaderboard
32 Claude Sonnet 4.5 173 Sep 2025 Official leaderboard
33 GPT-6 Luna 169 Sep 2026 Official leaderboard
34 GPT-5.6 Luna 164 Jul 2026 Official leaderboard
35 GPT-5 mini 154 Aug 2025 Official leaderboard
36 Grok 4.3 145 Apr 2026 Official leaderboard
37 GPT-5.4 Mini 140 Mar 2026 Official leaderboard
38 Grok 4.1 Fast 135 reasoning Nov 2025 Official leaderboard
39 DeepSeek-V3.2 0 Dec 2025 Official leaderboard

7 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 Anthropic
Own leaderboard 7 leaders
Gemini 4 Argon 535

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 2025 2026 0 200 400 600 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