Atlas

Benchmarks

← All benchmarks

OSWorld (self-reported, Set-of-Mark)

Agents · 2024-04-11

Self-reported OSWorld success rate with Set-of-Mark observation.

Top models (higher is better)

ModelScore
GPT-4o24.5
GPT-4 Turbo with Vision11.8
GPT-4 Turbo8.4
Gemini 1.5 Pro7.8
Opus 36.7
Qwen VL Max 08095.4
GPT-4o Mini3.5
LLaVA-OneVision 72B2.4
Gemini 1.0 Pro Vision1.1
CogAgent1.0
Loading Atlas data…