Benchmark Almanac

Game Arena Dark Hex

Gemini 3.8 Flash 612 leads since Sep 2026 Not in model cards

Language models play Hex without seeing the opponent's stones, so they must probe cells to learn the hidden board while building a path.

Kaggle published the leaderboard on 2026-06-04 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 Jun 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 3.8 Flash leads with 612

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. 612 Gemini 3.8 Flash
  2. 482 GPT-5.5
  3. 479 Gemini 4 Argon
  4. 438 Gemini 3.5 Flash
  5. 379 GPT-5.4
  6. 368 Claude Opus 5
  7. 348 Gemini 3 Flash
  8. 341 GPT-5.6 Sol
  9. 329 Gemini 3.6 Flash
  10. 296 Gemini 3.1 Pro
  11. 290 Gemini 3.7 Flash
  12. 288 Grok 4.6
  13. 283 Claude Opus 5.5
  14. 281 Claude Sonnet 5.5
  15. 280 GPT-6 Astra
  16. 280 GPT-6.1 Sol
  17. 258 Claude Opus 4.8
  18. 257 Claude Fable 5.1
  19. 255 Grok 4.7
  20. 247 GPT-5.2
  21. 246 GPT-6 Sol
  22. 239 Claude Haiku 5.5
  23. 221 o3
  24. 201 Claude Sonnet 5
  25. 187 GPT-5 mini
  26. 173 Grok 4.5
  27. 144 GPT-5.6 Luna
  28. 137 GPT-6 Luna
  29. 131 GPT-5.6 Terra
  30. 126 Grok 4.1 Fast
  31. 115 Claude Sonnet 4.6
  32. 109 Claude Opus 4.6
  33. 98 GPT-5.4 Mini
  34. 97 Claude Opus 4.7
  35. 70 Grok 4.3
  36. 45 Grok 4.20
  37. 42 Claude Sonnet 4.5
  38. 0 DeepSeek-V3.2
  1. 1 Gemini 3.8 Flash 612
  2. 2 GPT-5.5 482
  3. 3 Gemini 4 Argon 479
  4. 4 Gemini 3.5 Flash 438
  5. 5 GPT-5.4 379
  6. 6 Claude Opus 5 368
  7. 7 Gemini 3 Flash 348
  8. 8 GPT-5.6 Sol 341
  9. 9 Gemini 3.6 Flash 329
  10. 10 Gemini 3.1 Pro 296
  11. 11 Gemini 3.7 Flash 290
  12. 12 Grok 4.6 288
  13. 13 Claude Opus 5.5 283
  14. 14 Claude Sonnet 5.5 281
  15. 15 GPT-6 Astra 280
  16. 16 GPT-6.1 Sol 280
  17. 17 Claude Opus 4.8 258
  18. 18 Claude Fable 5.1 257
  19. 19 Grok 4.7 255
  20. 20 GPT-5.2 247
  21. 21 GPT-6 Sol 246
  22. 22 Claude Haiku 5.5 239
  23. 23 o3 221
  24. 24 Claude Sonnet 5 201
  25. 25 GPT-5 mini 187
  26. 26 Grok 4.5 173
  27. 27 GPT-5.6 Luna 144
  28. 28 GPT-6 Luna 137
  29. 29 GPT-5.6 Terra 131
  30. 30 Grok 4.1 Fast reasoning 126
  31. 31 Claude Sonnet 4.6 115
  32. 32 Claude Opus 4.6 109
  33. 33 GPT-5.4 Mini 98
  34. 34 Claude Opus 4.7 97
  35. 35 Grok 4.3 70
  36. 36 Grok 4.20 reasoning 45
  37. 37 Claude Sonnet 4.5 42
  38. 38 DeepSeek-V3.2 0
#ModelScoreSettingDateSource
1 Gemini 3.8 Flash 612 Sep 2026 Official leaderboard
2 GPT-5.5 482 Apr 2026 Official leaderboard
3 Gemini 4 Argon 479 Sep 2026 Official leaderboard
4 Gemini 3.5 Flash 438 May 2026 Official leaderboard
5 GPT-5.4 379 Mar 2026 Official leaderboard
6 Claude Opus 5 368 Jul 2026 Official leaderboard
7 Gemini 3 Flash 348 Dec 2025 Official leaderboard
8 GPT-5.6 Sol 341 Jul 2026 Official leaderboard
9 Gemini 3.6 Flash 329 Jul 2026 Official leaderboard
10 Gemini 3.1 Pro 296 Feb 2026 Official leaderboard
11 Gemini 3.7 Flash 290 Aug 2026 Official leaderboard
12 Grok 4.6 288 Aug 2026 Official leaderboard
13 Claude Opus 5.5 283 Sep 2026 Official leaderboard
14 Claude Sonnet 5.5 281 Sep 2026 Official leaderboard
15 GPT-6 Astra 280 Sep 2026 Official leaderboard
16 GPT-6.1 Sol 280 Sep 2026 Official leaderboard
17 Claude Opus 4.8 258 May 2026 Official leaderboard
18 Claude Fable 5.1 257 Sep 2026 Official leaderboard
19 Grok 4.7 255 Sep 2026 Official leaderboard
20 GPT-5.2 247 Dec 2025 Official leaderboard
21 GPT-6 Sol 246 Sep 2026 Official leaderboard
22 Claude Haiku 5.5 239 Oct 2026 Official leaderboard
23 o3 221 Apr 2025 Official leaderboard
24 Claude Sonnet 5 201 Jun 2026 Official leaderboard
25 GPT-5 mini 187 Aug 2025 Official leaderboard
26 Grok 4.5 173 Jul 2026 Official leaderboard
27 GPT-5.6 Luna 144 Jul 2026 Official leaderboard
28 GPT-6 Luna 137 Sep 2026 Official leaderboard
29 GPT-5.6 Terra 131 Jul 2026 Official leaderboard
30 Grok 4.1 Fast 126 reasoning Nov 2025 Official leaderboard
31 Claude Sonnet 4.6 115 Feb 2026 Official leaderboard
32 Claude Opus 4.6 109 Feb 2026 Official leaderboard
33 GPT-5.4 Mini 98 Mar 2026 Official leaderboard
34 Claude Opus 4.7 97 Apr 2026 Official leaderboard
35 Grok 4.3 70 Apr 2026 Official leaderboard
36 Grok 4.20 45 reasoning Feb 2026 Official leaderboard
37 Claude Sonnet 4.5 42 Sep 2025 Official leaderboard
38 DeepSeek-V3.2 0 Dec 2025 Official leaderboard

2 models have held number one since its release in Jun 2026

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

OpenAI Google DeepMind
Own leaderboard 2 leaders
Gemini 3.8 Flash 612

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 0 200 400 600 800 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