Atlas

Benchmarks

← All benchmarks

SciUniverse Level 1: Biology

Science

The 23 biology tasks of SciUniverse Level 1 in five families: DNA amplification and recovery and cell-free sfGFP expression in the physical lab, LC-MRM peak review, and simulated Hamilton STARlet PCR and normalization runs driven through VENUS, PyLabRobot or firmware commands. Pass@1 weights the five families equally.

Top models (higher is better)

ModelScore
GPT-6 Astra60.0
Claude Fable 5.140.0
Grok 4.632.0
Claude Opus 531.1
Gemini 3.8 Flash28.9
GPT-5.6 Sol15.6
Loading Atlas data…