Skip to content

Gert Labs Composite Game Benchmark

A game-environment benchmark that evaluates AI models in novel games covering strategic planning, resource management, spatial reasoning, cooperation, and theory of mind.

52 models

Higher is better.

  1. 1Claude Opus 4.8Anthropic72.97
  2. 2GPT-5.5OpenAI72.93
  3. 3Claude Opus 4.7Anthropic65.59
  4. 4GPT-5.4OpenAI64.89
  5. 5Qwen3.7 MaxAlibaba64.27
  6. 6Claude Opus 4.5Anthropic64.23
  7. 7Gemini 3 ProGoogle63.23
  8. 8Claude Sonnet 4.6Anthropic62.92
  9. 9MiMo-V2.5-ProXiaomi62.70
  10. 10Claude Opus 4.6Anthropic61.85
  11. 11Gemini 3.5 FlashGoogle61.85
  12. 12GLM-5.1Z.AI60.11
  13. 13GPT-5.3 CodexOpenAI57.47
  14. 14Gemini 3.1 ProGoogle56.87
  15. 15Kimi K2.6Moonshot AI56.82
  16. 16Gemini 3 FlashGoogle56.63
  17. 17Qwen3.6-27BAlibaba54.84
  18. 18GPT-5.2-CodexOpenAI51.79
  19. 19Step 3.7 FlashStepFun51.57
  20. 20GLM-5Z.AI50.99
  21. 21Qwen3.6 PlusAlibaba50.60
  22. 22GPT-5.1-CodexOpenAI49.68
  23. 23Grok Build 0.1xAI49.15
  24. 24Claude Sonnet 4.5Anthropic48.51
  25. 25Grok 4.1 FastxAI47.32
  26. 26MiMo-V2.5Xiaomi46.89
  27. 27Qwen3.5 397BAlibaba46.76
  28. 28GPT-5.2OpenAI46.54
  29. 29Kimi K2.5Moonshot AI45.88
  30. 30Grok 4.3xAI43.86
  31. 31Qwen3 MaxAlibaba43.74
  32. 32Qwen3.6-35B-A3BAlibaba42.65
  33. 33Grok 4xAI42.34
  34. 34Gemini 2.5 ProGoogle42.01
  35. 35GPT-5.1OpenAI41.24
  36. 36MiniMax M2.7MiniMax40.40
  37. 37GLM-4.7Z.AI39.95
  38. 38Claude 4 SonnetAnthropic39.66
  39. 39Qwen3.5-27BAlibaba39.41
  40. 40Mistral Medium 3.5 128BMistral39.10
Higher is better. Showing the top 40 of 52. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Composite game leaderboard
Tasks
Novel game environments
Authors
Gert Labs (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.

All results

1Claude Opus 4.8 Anthropic72.97
2GPT-5.5 OpenAI72.93
3Claude Opus 4.7 Anthropic65.59
4GPT-5.4 OpenAI64.89
5Qwen3.7 Max Alibaba64.27
6Claude Opus 4.5 Anthropic64.23
7Gemini 3 Pro Google63.23
8Claude Sonnet 4.6 Anthropic62.92
9MiMo-V2.5-Pro Xiaomi62.70
10Claude Opus 4.6 Anthropic61.85
11Gemini 3.5 Flash Google61.85
12GLM-5.1 Z.AI60.11
13GPT-5.3 Codex OpenAI57.47
14Gemini 3.1 Pro Google56.87
15Kimi K2.6 Moonshot AI56.82
16Gemini 3 Flash Google56.63
17Qwen3.6-27B Alibaba54.84
18GPT-5.2-Codex OpenAI51.79
19Step 3.7 Flash StepFun51.57
20GLM-5 Z.AI50.99
21Qwen3.6 Plus Alibaba50.60
22GPT-5.1-Codex OpenAI49.68
23Grok Build 0.1 xAI49.15
24Claude Sonnet 4.5 Anthropic48.51
25Grok 4.1 Fast xAI47.32
26MiMo-V2.5 Xiaomi46.89
27Qwen3.5 397B Alibaba46.76
28GPT-5.2 OpenAI46.54
29Kimi K2.5 Moonshot AI45.88
30Grok 4.3 xAI43.86
31Qwen3 Max Alibaba43.74
32Qwen3.6-35B-A3B Alibaba42.65
33Grok 4 xAI42.34
34Gemini 2.5 Pro Google42.01
35GPT-5.1 OpenAI41.24
36MiniMax M2.7 MiniMax40.40
37GLM-4.7 Z.AI39.95
38Claude 4 Sonnet Anthropic39.66
39Qwen3.5-27B Alibaba39.41
40Mistral Medium 3.5 128B Mistral39.10
41Gemini 3.1 Flash-Lite Google38.46
42Grok 4.20 xAI38.36
43Hy3 Preview Tencent36.91
44MiMo-V2-Pro Xiaomi36.68
45Gemma 4 31B Google35.26
46Kimi K2.5 (Reasoning) Moonshot AI32.58
47Trinity-Large-Thinking Arcee AI32.55
48GLM-5V-Turbo Z.AI30.76
49GPT-OSS 120B OpenAI29.61
50DeepSeek V3.2 DeepSeek29.57
51Qwen3.5-35B-A3B Alibaba28.96
52GPT-4.1 OpenAI25.65