Benchmark Almanac

Game Arena Word Association

Gemini 4 Carbon 734 leads since October 2026
Not in model cards

Teams of language models give one-word clues and guess linked words, in a game like Codenames, against an opposing team.

Kaggle published the leaderboard on May 26th, 2026 with 36 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
May 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
40
Models scored on its own leaderboard

Gemini 4 Carbon leads with 734

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.

40 models. Scroll sideways for the rest.

  1. 734 Gemini 4 Carbon
  2. 625 Gemini 4 Argon
  3. 563 GPT-5.5
  4. 503 GPT-6 Astra
  5. 444 Claude Fable 5.2
  6. 441 GPT-6.1 Sol
  7. 356 o3
  8. 350 GPT-5.6 Sol
  9. 344 Gemini 3.8 Flash
  10. 340 GPT-6 Sol
  11. 329 Gemini 3.7 Flash
  12. 316 Gemini 3.1 Pro
  13. 310 GPT-6.1 Terra
  14. 303 Claude Opus 4.6
  15. 295 Gemini 3.5 Flash
  16. 290 Claude Opus 5.5
  17. 289 Grok 4.6
  18. 288 Gemini 3.6 Flash
  19. 282 GPT-5.4
  20. 279 Grok 4.7
  21. 262 Gemini 3 Flash
  22. 254 Grok 4.5
  23. 216 Claude Sonnet 4.6
  24. 209 Claude Opus 4.7
  25. 188 GPT-5.2
  26. 177 GPT-5.6 Terra
  27. 160 Claude Opus 4.8
  28. 127 Claude Sonnet 4.5
  29. 126 Claude Sonnet 5.5
  30. 117 GPT-6 Luna
  31. 108 Grok 4.3
  32. 83 GPT-5 mini
  33. 82 GPT-5.6 Luna
  34. 64 GPT-5.4 Mini
  35. 58 Grok 4.20
  36. 37 Grok 4.1 Fast
  37. 25 Claude Haiku 5.5
  38. 25 Claude Sonnet 5
  39. 15 DeepSeek-V3.2
  40. 0 Grok 4
  1. 1 Gemini 4 Carbon 734
  2. 2 Gemini 4 Argon 625
  3. 3 GPT-5.5 563
  4. 4 GPT-6 Astra 503
  5. 5 Claude Fable 5.2 444
  6. 6 GPT-6.1 Sol 441
  7. 7 o3 356
  8. 8 GPT-5.6 Sol 350
  9. 9 Gemini 3.8 Flash 344
  10. 10 GPT-6 Sol 340
  11. 11 Gemini 3.7 Flash 329
  12. 12 Gemini 3.1 Pro 316
  13. 13 GPT-6.1 Terra 310
  14. 14 Claude Opus 4.6 303
  15. 15 Gemini 3.5 Flash 295
  16. 16 Claude Opus 5.5 290
  17. 17 Grok 4.6 289
  18. 18 Gemini 3.6 Flash 288
  19. 19 GPT-5.4 282
  20. 20 Grok 4.7 279
  21. 21 Gemini 3 Flash 262
  22. 22 Grok 4.5 254
  23. 23 Claude Sonnet 4.6 216
  24. 24 Claude Opus 4.7 209
  25. 25 GPT-5.2 188
  26. 26 GPT-5.6 Terra 177
  27. 27 Claude Opus 4.8 160
  28. 28 Claude Sonnet 4.5 127
  29. 29 Claude Sonnet 5.5 126
  30. 30 GPT-6 Luna 117
  31. 31 Grok 4.3 108
  32. 32 GPT-5 mini 83
  33. 33 GPT-5.6 Luna 82
  34. 34 GPT-5.4 Mini 64
  35. 35 Grok 4.20 reasoning 58
  36. 36 Grok 4.1 Fast reasoning 37
  37. 37 Claude Haiku 5.5 25
  38. 38 Claude Sonnet 5 25
  39. 39 DeepSeek-V3.2 15
  40. 40 Grok 4 0

Source: Official Game Arena Word Association leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 734 Oct 2026 Official leaderboard
2 Gemini 4 Argon 625 Sep 2026 Official leaderboard
3 GPT-5.5 563 Apr 2026 Official leaderboard
4 GPT-6 Astra 503 Sep 2026 Official leaderboard
5 Claude Fable 5.2 444 Nov 2026 Official leaderboard
6 GPT-6.1 Sol 441 Sep 2026 Official leaderboard
7 o3 356 Apr 2025 Official leaderboard
8 GPT-5.6 Sol 350 Jul 2026 Official leaderboard
9 Gemini 3.8 Flash 344 Sep 2026 Official leaderboard
10 GPT-6 Sol 340 Sep 2026 Official leaderboard
11 Gemini 3.7 Flash 329 Aug 2026 Official leaderboard
12 Gemini 3.1 Pro 316 Feb 2026 Official leaderboard
13 GPT-6.1 Terra 310 Oct 2026 Official leaderboard
14 Claude Opus 4.6 303 Feb 2026 Official leaderboard
15 Gemini 3.5 Flash 295 May 2026 Official leaderboard
16 Claude Opus 5.5 290 Sep 2026 Official leaderboard
17 Grok 4.6 289 Aug 2026 Official leaderboard
18 Gemini 3.6 Flash 288 Jul 2026 Official leaderboard
19 GPT-5.4 282 Mar 2026 Official leaderboard
20 Grok 4.7 279 Sep 2026 Official leaderboard
21 Gemini 3 Flash 262 Dec 2025 Official leaderboard
22 Grok 4.5 254 Jul 2026 Official leaderboard
23 Claude Sonnet 4.6 216 Feb 2026 Official leaderboard
24 Claude Opus 4.7 209 Apr 2026 Official leaderboard
25 GPT-5.2 188 Dec 2025 Official leaderboard
26 GPT-5.6 Terra 177 Jul 2026 Official leaderboard
27 Claude Opus 4.8 160 May 2026 Official leaderboard
28 Claude Sonnet 4.5 127 Sep 2025 Official leaderboard
29 Claude Sonnet 5.5 126 Sep 2026 Official leaderboard
30 GPT-6 Luna 117 Sep 2026 Official leaderboard
31 Grok 4.3 108 Apr 2026 Official leaderboard
32 GPT-5 mini 83 Aug 2025 Official leaderboard
33 GPT-5.6 Luna 82 Jul 2026 Official leaderboard
34 GPT-5.4 Mini 64 Mar 2026 Official leaderboard
35 Grok 4.20 58 reasoning Feb 2026 Official leaderboard
36 Grok 4.1 Fast 37 reasoning Nov 2025 Official leaderboard
37 Claude Haiku 5.5 25 Oct 2026 Official leaderboard
38 Claude Sonnet 5 25 Jun 2026 Official leaderboard
39 DeepSeek-V3.2 15 Dec 2025 Official leaderboard
40 Grok 4 0 Jul 2025 Official leaderboard

3 models have held number one on the official leaderboard since its release in May 2026

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

OpenAI Google DeepMind
Own leaderboard 3 leaders
Gemini 4 Carbon 734

Source: Official Game Arena Word Association 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 Game Arena Word Association 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%