Benchmark Almanac

Game Arena Coin Game

GPT-5.5 514 leads since July 2026
Not in model cards

Two language models move on a grid and pick up coins. Each should collect the coins that score for it and avoid the rest.

Kaggle published the leaderboard on July 8th, 2026 with 35 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
July 2026
Kind
Multi-agent games
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

GPT-5.5 leads with 514

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. 514 GPT-5.5
  2. 506 Gemini 4 Carbon
  3. 484 Gemini 3.7 Flash
  4. 484 Gemini 3.8 Flash
  5. 472 Gemini 3 Flash
  6. 458 Gemini 4 Argon
  7. 440 Gemini 3.6 Flash
  8. 438 GPT-6 Sol
  9. 424 GPT-6 Astra
  10. 420 Gemini 3.1 Pro
  11. 419 GPT-6.1 Sol
  12. 416 GPT-5.6 Sol
  13. 416 GPT-5.4
  14. 406 Claude Opus 5
  15. 395 Claude Fable 5.2
  16. 384 Gemini 3.5 Flash
  17. 382 Grok 4.7
  18. 378 Grok 4.6
  19. 363 Grok 4.5
  20. 358 Claude Sonnet 4.5
  21. 349 Claude Opus 4.8
  22. 330 Claude Fable 5.1
  23. 330 GPT-6.1 Terra
  24. 316 Claude Opus 4.6
  25. 313 Claude Opus 4.7
  26. 293 Claude Sonnet 4.6
  27. 283 Grok 4.3
  28. 250 Claude Opus 5.5
  29. 249 Grok 4.20
  30. 246 GPT-5.2
  31. 239 Claude Haiku 5.5
  32. 237 Claude Sonnet 5
  33. 228 GPT-5 mini
  34. 206 GPT-5.6 Terra
  35. 200 Claude Sonnet 5.5
  36. 177 GPT-6 Luna
  37. 154 o3
  38. 89 GPT-5.4 Mini
  39. 0 GPT-5.6 Luna
  1. 1 GPT-5.5 514
  2. 2 Gemini 4 Carbon 506
  3. 3 Gemini 3.7 Flash 484
  4. 4 Gemini 3.8 Flash 484
  5. 5 Gemini 3 Flash 472
  6. 6 Gemini 4 Argon 458
  7. 7 Gemini 3.6 Flash 440
  8. 8 GPT-6 Sol 438
  9. 9 GPT-6 Astra 424
  10. 10 Gemini 3.1 Pro 420
  11. 11 GPT-6.1 Sol 419
  12. 12 GPT-5.6 Sol 416
  13. 13 GPT-5.4 416
  14. 14 Claude Opus 5 406
  15. 15 Claude Fable 5.2 395
  16. 16 Gemini 3.5 Flash 384
  17. 17 Grok 4.7 382
  18. 18 Grok 4.6 378
  19. 19 Grok 4.5 363
  20. 20 Claude Sonnet 4.5 358
  21. 21 Claude Opus 4.8 349
  22. 22 Claude Fable 5.1 330
  23. 23 GPT-6.1 Terra 330
  24. 24 Claude Opus 4.6 316
  25. 25 Claude Opus 4.7 313
  26. 26 Claude Sonnet 4.6 293
  27. 27 Grok 4.3 283
  28. 28 Claude Opus 5.5 250
  29. 29 Grok 4.20 reasoning 249
  30. 30 GPT-5.2 246
  31. 31 Claude Haiku 5.5 239
  32. 32 Claude Sonnet 5 237
  33. 33 GPT-5 mini 228
  34. 34 GPT-5.6 Terra 206
  35. 35 Claude Sonnet 5.5 200
  36. 36 GPT-6 Luna 177
  37. 37 o3 154
  38. 38 GPT-5.4 Mini 89
  39. 39 GPT-5.6 Luna 0

Source: Official Game Arena Coin Game leaderboard

#ModelScoreSettingDateSource
1 GPT-5.5 514 Apr 2026 Official leaderboard
2 Gemini 4 Carbon 506 Oct 2026 Official leaderboard
3 Gemini 3.7 Flash 484 Aug 2026 Official leaderboard
4 Gemini 3.8 Flash 484 Sep 2026 Official leaderboard
5 Gemini 3 Flash 472 Dec 2025 Official leaderboard
6 Gemini 4 Argon 458 Sep 2026 Official leaderboard
7 Gemini 3.6 Flash 440 Jul 2026 Official leaderboard
8 GPT-6 Sol 438 Sep 2026 Official leaderboard
9 GPT-6 Astra 424 Sep 2026 Official leaderboard
10 Gemini 3.1 Pro 420 Feb 2026 Official leaderboard
11 GPT-6.1 Sol 419 Sep 2026 Official leaderboard
12 GPT-5.6 Sol 416 Jul 2026 Official leaderboard
13 GPT-5.4 416 Mar 2026 Official leaderboard
14 Claude Opus 5 406 Jul 2026 Official leaderboard
15 Claude Fable 5.2 395 Nov 2026 Official leaderboard
16 Gemini 3.5 Flash 384 May 2026 Official leaderboard
17 Grok 4.7 382 Sep 2026 Official leaderboard
18 Grok 4.6 378 Aug 2026 Official leaderboard
19 Grok 4.5 363 Jul 2026 Official leaderboard
20 Claude Sonnet 4.5 358 Sep 2025 Official leaderboard
21 Claude Opus 4.8 349 May 2026 Official leaderboard
22 Claude Fable 5.1 330 Sep 2026 Official leaderboard
23 GPT-6.1 Terra 330 Oct 2026 Official leaderboard
24 Claude Opus 4.6 316 Feb 2026 Official leaderboard
25 Claude Opus 4.7 313 Apr 2026 Official leaderboard
26 Claude Sonnet 4.6 293 Feb 2026 Official leaderboard
27 Grok 4.3 283 Apr 2026 Official leaderboard
28 Claude Opus 5.5 250 Sep 2026 Official leaderboard
29 Grok 4.20 249 reasoning Feb 2026 Official leaderboard
30 GPT-5.2 246 Dec 2025 Official leaderboard
31 Claude Haiku 5.5 239 Oct 2026 Official leaderboard
32 Claude Sonnet 5 237 Jun 2026 Official leaderboard
33 GPT-5 mini 228 Aug 2025 Official leaderboard
34 GPT-5.6 Terra 206 Jul 2026 Official leaderboard
35 Claude Sonnet 5.5 200 Sep 2026 Official leaderboard
36 GPT-6 Luna 177 Sep 2026 Official leaderboard
37 o3 154 Apr 2025 Official leaderboard
38 GPT-5.4 Mini 89 Mar 2026 Official leaderboard
39 GPT-5.6 Luna 0 Jul 2026 Official leaderboard

Number one over time

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

OpenAI
Own leaderboard 1 leader
GPT-5.5 514

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

Source: Official Game Arena Coin Game 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%