Atlas

Benchmarks

← All benchmarks

Robo-bench — Balance Scale (public)

Agents · 2026-10-13

Public Robo-bench arm task: balance a beam scale against a sealed load of unknown mass with labelled weights. Milestones: weight on the right pan 10, beam off its stop 30, within 5 degrees 50, within 2 degrees 75, balanced and held for 2 s 100. A single public, single-seed episode nested in the arm suite, so it is a diagnostic.

Top models (higher is better)

ModelScore
GPT-6.1 Sol100.0
Claude Opus 5.595.0
Claude Fable 5.110.0
Muse Wonder Full10.0
Gemini 4 Argon0.0
GPT-6 Astra0.0
Grok 4.70.0
Loading Atlas data…