Benchmark Almanac

Game Arena Five in a Row

Gemini 3.1 Pro 1,131 leads since Apr 2026 Not in model cards

Language models play a gomoku-style game and try to line up five of their tokens before the opponent does.

Kaggle published the leaderboard on 2026-04-30 with 20 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 Apr 2026 Puzzles and chess All tasks public Last model added Jun 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
20
Models scored on its own leaderboard

Gemini 3.1 Pro leads with 1,131 among models released by Jun 2026

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.

The newest model with a score here came out in Jun 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

20 models. Scroll sideways for the rest.

  1. 1,131 Gemini 3.1 Pro
  2. 925 GPT-5.5
  3. 878 GPT-5.4
  4. 865 Claude Sonnet 5
  5. 817 GPT-5.2
  6. 786 Gemini 3.5 Flash
  7. 756 Gemini 3 Flash
  8. 618 Grok 4.20
  9. 540 o3
  10. 497 GPT-5.4 Mini
  11. 452 Claude Opus 4.8
  12. 442 Grok 4
  13. 387 GPT-5 mini
  14. 373 Grok 4.1 Fast
  15. 367 Grok 4.3
  16. 353 Claude Opus 4.7
  17. 315 Claude Opus 4.6
  18. 241 Claude Sonnet 4.6
  19. 126 DeepSeek-V3.2
  20. 0 Claude Sonnet 4.5
  1. 1 Gemini 3.1 Pro 1,131
  2. 2 GPT-5.5 925
  3. 3 GPT-5.4 878
  4. 4 Claude Sonnet 5 865
  5. 5 GPT-5.2 817
  6. 6 Gemini 3.5 Flash 786
  7. 7 Gemini 3 Flash 756
  8. 8 Grok 4.20 reasoning 618
  9. 9 o3 540
  10. 10 GPT-5.4 Mini 497
  11. 11 Claude Opus 4.8 452
  12. 12 Grok 4 442
  13. 13 GPT-5 mini 387
  14. 14 Grok 4.1 Fast reasoning 373
  15. 15 Grok 4.3 367
  16. 16 Claude Opus 4.7 353
  17. 17 Claude Opus 4.6 315
  18. 18 Claude Sonnet 4.6 241
  19. 19 DeepSeek-V3.2 126
  20. 20 Claude Sonnet 4.5 0
#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 1,131 Feb 2026 Official leaderboard
2 GPT-5.5 925 Apr 2026 Official leaderboard
3 GPT-5.4 878 Mar 2026 Official leaderboard
4 Claude Sonnet 5 865 Jun 2026 Official leaderboard
5 GPT-5.2 817 Dec 2025 Official leaderboard
6 Gemini 3.5 Flash 786 May 2026 Official leaderboard
7 Gemini 3 Flash 756 Dec 2025 Official leaderboard
8 Grok 4.20 618 reasoning Feb 2026 Official leaderboard
9 o3 540 Apr 2025 Official leaderboard
10 GPT-5.4 Mini 497 Mar 2026 Official leaderboard
11 Claude Opus 4.8 452 May 2026 Official leaderboard
12 Grok 4 442 Jul 2025 Official leaderboard
13 GPT-5 mini 387 Aug 2025 Official leaderboard
14 Grok 4.1 Fast 373 reasoning Nov 2025 Official leaderboard
15 Grok 4.3 367 Apr 2026 Official leaderboard
16 Claude Opus 4.7 353 Apr 2026 Official leaderboard
17 Claude Opus 4.6 315 Feb 2026 Official leaderboard
18 Claude Sonnet 4.6 241 Feb 2026 Official leaderboard
19 DeepSeek-V3.2 126 Dec 2025 Official leaderboard
20 Claude Sonnet 4.5 0 Sep 2025 Official leaderboard

Number one over time

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

Google DeepMind
Own leaderboard Not tested since Jun 2026
Gemini 3.1 Pro 1,131

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
0 250 500 750 1,000 1,250 2025 2026 0 250 500 750 1,000 1,250 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