Benchmark Almanac

Game Arena Bargaining

GPT-5.5 277 leads since Jul 2026 Not in model cards

Two language models negotiate how to split a pool of items, each with private item values. Each side tries to reach a good deal without giving away its values.

Kaggle published the leaderboard on 2026-07-23 with 36 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 Jul 2026 Multi-agent games 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
36
Models scored on its own leaderboard

GPT-5.5 leads with 277

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.

36 models. Scroll sideways for the rest.

  1. 277 GPT-5.5
  2. 250 GPT-6 Astra
  3. 248 Gemini 4 Argon
  4. 247 Claude Sonnet 5.5
  5. 245 GPT-6.1 Sol
  6. 239 Claude Opus 5.5
  7. 230 Claude Fable 5.1
  8. 226 Grok 4.20
  9. 224 GPT-5.6 Sol
  10. 223 Grok 4.5
  11. 220 Gemini 3.7 Flash
  12. 218 Grok 4.6
  13. 214 Gemini 3.8 Flash
  14. 210 GPT-6 Sol
  15. 196 Gemini 3.1 Pro
  16. 195 Claude Opus 4.7
  17. 193 o3
  18. 192 GPT-5.4
  19. 186 Gemini 3 Flash
  20. 178 Claude Opus 4.6
  21. 178 Claude Opus 4.8
  22. 178 Gemini 3.6 Flash
  23. 170 Grok 4.7
  24. 167 Claude Sonnet 4.5
  25. 165 Claude Haiku 5.5
  26. 162 GPT-5.6 Luna
  27. 158 Claude Opus 5
  28. 148 GPT-5 mini
  29. 148 GPT-6 Luna
  30. 147 Gemini 3.5 Flash
  31. 136 GPT-5.6 Terra
  32. 131 GPT-5.2
  33. 131 Grok 4.3
  34. 87 Claude Sonnet 5
  35. 78 Claude Sonnet 4.6
  36. 0 GPT-5.4 Mini
  1. 1 GPT-5.5 277
  2. 2 GPT-6 Astra 250
  3. 3 Gemini 4 Argon 248
  4. 4 Claude Sonnet 5.5 247
  5. 5 GPT-6.1 Sol 245
  6. 6 Claude Opus 5.5 239
  7. 7 Claude Fable 5.1 230
  8. 8 Grok 4.20 reasoning 226
  9. 9 GPT-5.6 Sol 224
  10. 10 Grok 4.5 223
  11. 11 Gemini 3.7 Flash 220
  12. 12 Grok 4.6 218
  13. 13 Gemini 3.8 Flash 214
  14. 14 GPT-6 Sol 210
  15. 15 Gemini 3.1 Pro 196
  16. 16 Claude Opus 4.7 195
  17. 17 o3 193
  18. 18 GPT-5.4 192
  19. 19 Gemini 3 Flash 186
  20. 20 Claude Opus 4.6 178
  21. 21 Claude Opus 4.8 178
  22. 22 Gemini 3.6 Flash 178
  23. 23 Grok 4.7 170
  24. 24 Claude Sonnet 4.5 167
  25. 25 Claude Haiku 5.5 165
  26. 26 GPT-5.6 Luna 162
  27. 27 Claude Opus 5 158
  28. 28 GPT-5 mini 148
  29. 29 GPT-6 Luna 148
  30. 30 Gemini 3.5 Flash 147
  31. 31 GPT-5.6 Terra 136
  32. 32 GPT-5.2 131
  33. 33 Grok 4.3 131
  34. 34 Claude Sonnet 5 87
  35. 35 Claude Sonnet 4.6 78
  36. 36 GPT-5.4 Mini 0
#ModelScoreSettingDateSource
1 GPT-5.5 277 Apr 2026 Official leaderboard
2 GPT-6 Astra 250 Sep 2026 Official leaderboard
3 Gemini 4 Argon 248 Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 247 Sep 2026 Official leaderboard
5 GPT-6.1 Sol 245 Sep 2026 Official leaderboard
6 Claude Opus 5.5 239 Sep 2026 Official leaderboard
7 Claude Fable 5.1 230 Sep 2026 Official leaderboard
8 Grok 4.20 226 reasoning Feb 2026 Official leaderboard
9 GPT-5.6 Sol 224 Jul 2026 Official leaderboard
10 Grok 4.5 223 Jul 2026 Official leaderboard
11 Gemini 3.7 Flash 220 Aug 2026 Official leaderboard
12 Grok 4.6 218 Aug 2026 Official leaderboard
13 Gemini 3.8 Flash 214 Sep 2026 Official leaderboard
14 GPT-6 Sol 210 Sep 2026 Official leaderboard
15 Gemini 3.1 Pro 196 Feb 2026 Official leaderboard
16 Claude Opus 4.7 195 Apr 2026 Official leaderboard
17 o3 193 Apr 2025 Official leaderboard
18 GPT-5.4 192 Mar 2026 Official leaderboard
19 Gemini 3 Flash 186 Dec 2025 Official leaderboard
20 Claude Opus 4.6 178 Feb 2026 Official leaderboard
21 Claude Opus 4.8 178 May 2026 Official leaderboard
22 Gemini 3.6 Flash 178 Jul 2026 Official leaderboard
23 Grok 4.7 170 Sep 2026 Official leaderboard
24 Claude Sonnet 4.5 167 Sep 2025 Official leaderboard
25 Claude Haiku 5.5 165 Oct 2026 Official leaderboard
26 GPT-5.6 Luna 162 Jul 2026 Official leaderboard
27 Claude Opus 5 158 Jul 2026 Official leaderboard
28 GPT-5 mini 148 Aug 2025 Official leaderboard
29 GPT-6 Luna 148 Sep 2026 Official leaderboard
30 Gemini 3.5 Flash 147 May 2026 Official leaderboard
31 GPT-5.6 Terra 136 Jul 2026 Official leaderboard
32 GPT-5.2 131 Dec 2025 Official leaderboard
33 Grok 4.3 131 Apr 2026 Official leaderboard
34 Claude Sonnet 5 87 Jun 2026 Official leaderboard
35 Claude Sonnet 4.6 78 Feb 2026 Official leaderboard
36 GPT-5.4 Mini 0 Mar 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-5.5 277

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
0 100 200 300 2025 2026 0 100 200 300 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