Gert Labs

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.

In index
Agentic
Source ↗

Models52

Top result73.0%Claude Opus 4.8

Top-3 spread7.4 pts

Year2026

Result and price

Results

52 results

#ModelResult

Claude Opus 4.873.0%

GPT-5.572.9%

Claude Opus 4.765.6%

GPT-5.464.9%

Qwen3.7 Max64.3%

Claude Opus 4.564.2%

Gemini 3 Pro63.2%

Claude Sonnet 4.662.9%

MiMo-V2.5-Pro62.7%

Gemini 3.5 Flash61.9%

Claude Opus 4.661.9%

GLM-5.160.1%

GPT-5.3 Codex57.5%

Gemini 3.1 Pro56.9%

15MAKimi K2.656.8%

Gemini 3 Flash56.6%

Qwen3.6-27B54.8%

GPT-5.2-Codex51.8%

19STStep 3.7 Flash51.6%

GLM-551.0%

Qwen3.6 Plus50.6%

GPT-5.1-Codex49.7%

Grok Build 0.149.1%

Claude Sonnet 4.548.5%

Grok 4.1 Fast47.3%

MiMo-V2.546.9%

Qwen3.5 397B46.8%

GPT-5.246.5%

29MAKimi K2.545.9%

Grok 4.343.9%

Qwen3 Max43.7%

Qwen3.6-35B-A3B42.6%

Grok 442.3%

Gemini 2.5 Pro42.0%

GPT-5.141.2%

MiniMax M2.740.4%

GLM-4.740.0%

Claude 4 Sonnet39.7%

Qwen3.5-27B39.4%

Mistral Medium 3.5 128B39.1%

Gemini 3.1 Flash-Lite38.5%

Grok 4.2038.4%

43TEHy3 Preview36.9%

MiMo-V2-Pro36.7%

Gemma 4 31B35.3%

46MAKimi K2.5 (Reasoning)32.6%

47AATrinity-Large-Thinking32.5%

GLM-5V-Turbo30.8%

GPT-OSS 120B29.6%

DeepSeek V3.229.6%

Qwen3.5-35B-A3B29.0%

GPT-4.125.6%