Atlas

Benchmarks

← All benchmarks

SciUniverse Level 1: Cross-domain

Science

SciUniverse Level 1's single cross-domain task: managing a simulated automated PCR lab over two weeks through supply shortages, equipment failures and deadlines, scored across 16 cases. A single task cannot stand as a domain component, so this view is a diagnostic.

Top models (higher is better)

ModelScore
Claude Fable 5.134.4
Claude Opus 56.3
GPT-6 Astra6.3
Gemini 3.8 Flash0.0
GPT-5.6 Sol0.0
Grok 4.60.0
Loading Atlas data…