Atlas

Benchmarks

← All benchmarks

OSWorld (self-reported, Screenshot + A11y tree)

Agents · 2024-04-11

Self-reported OSWorld success rate with screenshot and accessibility-tree observation.

Top models (higher is better)

ModelScore
GPT-4o23.9
Claude 3.5 Sonnet (June 2024)20.5
GPT-4 Turbo with Vision12.2
GPT-4 Turbo9.0
GPT-4o Mini6.2
Gemini 1.5 Pro5.1
Opus 34.4
Qwen VL Max 08093.5
Gemini 1.0 Pro Vision3.5
LLaVA-OneVision 72B2.4
CogAgent1.3
Loading Atlas data…