OSWorld
Computer use by an agent operating a desktop from screenshots
agents and tool use
Definition
- Version
- not in the notes
- Category
- agents and tool use
- Maintainer
- not in the notes
- Status
- not in the notes
- Home page
- not in the notes
Sources
- anthropic.com/news/3-5-models-and-computer-use https://www.anthropic.com/news/3-5-models-and-computer-use
Scores
| Model | Score | Who ran it | Setting | View or population | Date | Source |
|---|---|---|---|---|---|---|
 Claude 3.5 Sonnet (Oct '24) | 22 % | vendor-reported | screenshot-only, more steps allowed | – | 22 Oct 2024 | anthropic.com |
| Claude 3.5 Sonnet (Oct '24) | 14.9 % | vendor-reported | screenshot-only | – | 22 Oct 2024 | anthropic.com |
Where this comes from
Scores from different settings, harnesses or dates are not directly comparable; the row says which. Sorted by score, highest first, as a reading aid, not a ranking.
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "OSWorld", https://aurion.semai.id/benchmarks/osworld/, records of October 2026.