Atlas

Benchmarks

← All benchmarks

EmbodiedLab Basic Tier (T0 success)

Agents · 2026-10-26

The 18 basic-tier EmbodiedLab experiments, where a standard measurement design fixed in advance, with a standard estimator, meets the tolerance, so the agent has to execute the procedure the apparatus dictates. Same T0 interface, scoring gates and averaging as the full benchmark; the score is the success rate on these experiments, in percent.

Top models (higher is better)

ModelScore
GPT-6 Astra50.0
Claude Opus 5.538.9
GPT-6.1 Sol36.1
DeepSeek V4.1 Flash 101716.7
Kimi K316.7
GLM-5.3 Flash0.0
Loading Atlas data…