Benchmark Almanac

Game Arena Four in a Row

Gemini 4 Argon 923 leads since Sep 2026 Not in model cards

Language models play a Connect Four style game and try to line up four of their tokens before the opponent does.

Kaggle published the leaderboard on 2026-02-19 with 38 models on 2026-11-01. Gameplay logs are a public Kaggle dataset. Part of the Kaggle Game Arena suite.

Made by Kaggle and Google DeepMind Released Feb 2026 Puzzles and chess All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Everyday Most adults, without training
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.
Kaggle/kaggle-environments 459 stars Dataset Leaderboard kaggle.com
38
Models scored on its own leaderboard

Gemini 4 Argon leads with 923

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.

38 models. Scroll sideways for the rest.

  1. 923 Gemini 4 Argon
  2. 691 Gemini 3.8 Flash
  3. 690 GPT-6 Astra
  4. 672 Claude Opus 5
  5. 597 Claude Fable 5.1
  6. 589 Gemini 3.1 Pro
  7. 573 GPT-6.1 Sol
  8. 551 Gemini 3.5 Flash
  9. 544 Gemini 3.7 Flash
  10. 496 Gemini 3.6 Flash
  11. 493 Grok 4.6
  12. 476 GPT-5.5
  13. 471 Grok 4.7
  14. 435 GPT-5.4
  15. 423 Claude Opus 4.8
  16. 416 Claude Haiku 5.5
  17. 393 GPT-5.2
  18. 367 Grok 4.5
  19. 358 GPT-5.6 Sol
  20. 341 GPT-6 Sol
  21. 330 Claude Sonnet 5
  22. 302 Claude Opus 4.5
  23. 294 o3
  24. 291 Grok 4.20
  25. 265 Grok 4
  26. 256 Gemini 3 Flash
  27. 210 Claude Opus 4.6
  28. 206 GPT-5.4 Mini
  29. 199 Grok 4.1 Fast
  30. 197 Grok 4.3
  31. 193 Claude Opus 4.7
  32. 189 GPT-5.6 Luna
  33. 157 Claude Sonnet 4.5
  34. 136 Claude Sonnet 4.6
  35. 127 GPT-6 Luna
  36. 91 GPT-5.6 Terra
  37. 9 DeepSeek-V3.2
  38. 0 GPT-5 mini
  1. 1 Gemini 4 Argon 923
  2. 2 Gemini 3.8 Flash 691
  3. 3 GPT-6 Astra 690
  4. 4 Claude Opus 5 672
  5. 5 Claude Fable 5.1 597
  6. 6 Gemini 3.1 Pro 589
  7. 7 GPT-6.1 Sol 573
  8. 8 Gemini 3.5 Flash 551
  9. 9 Gemini 3.7 Flash 544
  10. 10 Gemini 3.6 Flash 496
  11. 11 Grok 4.6 493
  12. 12 GPT-5.5 476
  13. 13 Grok 4.7 471
  14. 14 GPT-5.4 435
  15. 15 Claude Opus 4.8 423
  16. 16 Claude Haiku 5.5 416
  17. 17 GPT-5.2 393
  18. 18 Grok 4.5 367
  19. 19 GPT-5.6 Sol 358
  20. 20 GPT-6 Sol 341
  21. 21 Claude Sonnet 5 330
  22. 22 Claude Opus 4.5 302
  23. 23 o3 294
  24. 24 Grok 4.20 reasoning 291
  25. 25 Grok 4 265
  26. 26 Gemini 3 Flash 256
  27. 27 Claude Opus 4.6 210
  28. 28 GPT-5.4 Mini 206
  29. 29 Grok 4.1 Fast reasoning 199
  30. 30 Grok 4.3 197
  31. 31 Claude Opus 4.7 193
  32. 32 GPT-5.6 Luna 189
  33. 33 Claude Sonnet 4.5 157
  34. 34 Claude Sonnet 4.6 136
  35. 35 GPT-6 Luna 127
  36. 36 GPT-5.6 Terra 91
  37. 37 DeepSeek-V3.2 9
  38. 38 GPT-5 mini 0
#ModelScoreSettingDateSource
1 Gemini 4 Argon 923 Sep 2026 Official leaderboard
2 Gemini 3.8 Flash 691 Sep 2026 Official leaderboard
3 GPT-6 Astra 690 Sep 2026 Official leaderboard
4 Claude Opus 5 672 Jul 2026 Official leaderboard
5 Claude Fable 5.1 597 Sep 2026 Official leaderboard
6 Gemini 3.1 Pro 589 Feb 2026 Official leaderboard
7 GPT-6.1 Sol 573 Sep 2026 Official leaderboard
8 Gemini 3.5 Flash 551 May 2026 Official leaderboard
9 Gemini 3.7 Flash 544 Aug 2026 Official leaderboard
10 Gemini 3.6 Flash 496 Jul 2026 Official leaderboard
11 Grok 4.6 493 Aug 2026 Official leaderboard
12 GPT-5.5 476 Apr 2026 Official leaderboard
13 Grok 4.7 471 Sep 2026 Official leaderboard
14 GPT-5.4 435 Mar 2026 Official leaderboard
15 Claude Opus 4.8 423 May 2026 Official leaderboard
16 Claude Haiku 5.5 416 Oct 2026 Official leaderboard
17 GPT-5.2 393 Dec 2025 Official leaderboard
18 Grok 4.5 367 Jul 2026 Official leaderboard
19 GPT-5.6 Sol 358 Jul 2026 Official leaderboard
20 GPT-6 Sol 341 Sep 2026 Official leaderboard
21 Claude Sonnet 5 330 Jun 2026 Official leaderboard
22 Claude Opus 4.5 302 Nov 2025 Official leaderboard
23 o3 294 Apr 2025 Official leaderboard
24 Grok 4.20 291 reasoning Feb 2026 Official leaderboard
25 Grok 4 265 Jul 2025 Official leaderboard
26 Gemini 3 Flash 256 Dec 2025 Official leaderboard
27 Claude Opus 4.6 210 Feb 2026 Official leaderboard
28 GPT-5.4 Mini 206 Mar 2026 Official leaderboard
29 Grok 4.1 Fast 199 reasoning Nov 2025 Official leaderboard
30 Grok 4.3 197 Apr 2026 Official leaderboard
31 Claude Opus 4.7 193 Apr 2026 Official leaderboard
32 GPT-5.6 Luna 189 Jul 2026 Official leaderboard
33 Claude Sonnet 4.5 157 Sep 2025 Official leaderboard
34 Claude Sonnet 4.6 136 Feb 2026 Official leaderboard
35 GPT-6 Luna 127 Sep 2026 Official leaderboard
36 GPT-5.6 Terra 91 Jul 2026 Official leaderboard
37 DeepSeek-V3.2 9 Dec 2025 Official leaderboard
38 GPT-5 mini 0 Aug 2025 Official leaderboard

5 models have held number one since its release in Feb 2026

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

OpenAI Google DeepMind Anthropic
Own leaderboard 5 leaders
Gemini 4 Argon 923

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 0 200 400 600 800 1,000 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