Atlas

Benchmarks

← All benchmarks

Butter-Bench: End-to-End Pass the Butter

Agents · 2025-10-23

Butter-Bench end-to-end task measuring the full butter delivery sequence from kitchen navigation through delivery and return to dock.

Top models (higher is better)

ModelScore
Human (Butter-Bench)100.0
Gemini 2.5 Pro40.0
Opus 4.140.0
GPT-540.0
Grok 440.0
Gemini Robotics-ER 1.50.0
Llama 4 Maverick0.0
Loading Atlas data…