Atlas

Benchmarks

← All benchmarks

GeneBench-Pro Artificial Analysis Subset

Science · 2026-06-30

GeneBench-Pro Artificial Analysis subset reports pass rates on the 50 held-out problems provided to Artificial Analysis for independent third-party reporting.

Top models (higher is better)

ModelScore
GPT-5.6 Sol Pro18.0
GPT-5.6 Terra Pro17.6
GPT-5.6 Luna Pro16.4
GPT-5.6 Sol16.3
GPT-5.6 Terra16.2
GPT-5.6 Luna9.6
GPT-5.4 Pro9.2
GPT-5.5 Pro9.2
Gemini 3.5 Flash6.0
Opus 4.84.8
GPT-5.44.7
GPT-5.54.6
GLM-5.23.4
Qwen3.7-Max3.2
GPT-5.23.2
GPT-5.2 Pro2.4
DeepSeek-V4-Pro2.3
MiMo-V2.5-Pro1.8
Kimi K2.7 Code1.8
Qwen3.7-Plus1.8
DeepSeek-V4-Flash1.2
Gemini 3.1 Pro Preview1.0
MiniMax M30.7
GLM-5.10.7
Grok 4.30.6
Kimi K2.60.6
Hy3 preview0.4
MiMo-V2.50.4
MiniMax M2.70.0
Loading Atlas data…