Atlas

Models

← All models

Human (Butter-Bench)

Andon Labs · 2025-10-23

Human baseline reported by Andon Labs for Butter-Bench. It represents human task-completion performance on the physical robot benchmark and is stored only to preserve the benchmark's published baseline row.

Benchmark scores

BenchmarkScore
Butter-Bench95.0
Butter-Bench: End-to-End Pass the Butter100.0
Butter-Bench: Infer Butter Bag100.0
Butter-Bench: Multi-Step Spatial Path Planning100.0
Butter-Bench: Notice Absence100.0
Butter-Bench: Search for Package100.0
Butter-Bench: Wait for Confirmed Pick Up66.0
Loading Atlas data…