Atlas

Benchmarks

← All benchmarks

OSWorld 2.0 150-step Binary Accuracy

Agents · 2026-06-25

OSWorld 2.0 official leaderboard binary completion accuracy under a 150-step action budget over the 108-task v2 suite.

Top models (higher is better)

ModelScore
GPT-5.513.0
Opus 4.74.6
Sonnet 4.64.6
Kimi K2.61.9
MiniMax M31.9
Loading Atlas data…