ModelRefs / OSWorld Leaderboard
OSWorld Leaderboard
Computer-use agents across real OS apps (Ubuntu/Windows/macOS). Current leaders, methodology, and citation sources for OSWorld.
Overview
Computer-use agents across real OS apps (Ubuntu/Windows/macOS).
How it is measured: 369 tasks; success rate with screenshot loop.
How this benchmark is scored
| Category | agents |
|---|---|
| Maximum score | 100 % success |
| Direction | Higher is better |
Primary source: https://os-world.github.io/
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to OSWorld Leaderboard.
Frequently asked questions
What does the OSWorld benchmark measure?
Computer-use agents across real OS apps (Ubuntu/Windows/macOS).
What is the methodology?
369 tasks; success rate with screenshot loop.
Which AI model is the current OSWorld leader?
Claude Sonnet 4.5 (Anthropic) currently leads with 61.4 % success.