Atlas

Benchmarks

← All benchmarks

GeneBench-Pro Public Release Subset

Science · 2026-06-30

GeneBench-Pro public release subset reports pass rates on the 10 final public case-study problems from the 129-problem GeneBench-Pro suite.

Top models (higher is better)

ModelScore
GPT-5.6 Terra Pro36.0
GPT-5.6 Sol Pro34.0
GPT-5.6 Sol24.0
GPT-5.6 Luna Pro22.0
GPT-5.4 Pro20.0
Opus 4.818.0
GPT-5.5 Pro14.0
GPT-5.6 Luna9.0
GLM-5.28.9
GPT-5.48.0
GPT-5.58.0
GPT-5.6 Terra8.0
GPT-5.2 Pro8.0
Kimi K2.67.7
Gemini 3.1 Pro Preview7.0
Qwen3.7-Max6.4
MiMo-V2.5-Pro6.0
DeepSeek-V4-Flash5.6
Gemini 3.5 Flash5.0
Qwen3.7-Plus4.4
MiniMax M34.0
GPT-5.24.0
MiMo-V2.53.0
DeepSeek-V4-Pro2.0
Hy3 preview1.0
MiniMax M2.70.0
GLM-5.10.0
Grok 4.30.0
Kimi K2.7 Code0.0
Loading Atlas data…