OS-Atlas
Foundation action model for generalist GUI agents, open corpus and weights.
About
OS-Atlas was introduced on October 30, 2024 (arXiv 2410.23218) by researchers from Shanghai AI Lab, SJTU, HKU, and MIT, and later presented as an ICLR 2025 spotlight. It is a foundation action model for GUI agents: alongside the model, the team released an open-source toolkit for synthesizing GUI grounding data and the largest open cross-platform GUI grounding corpus to date, with over 13 million GUI elements spanning Windows, Linux, macOS, Android, and the web.
A unified action space resolves action-naming inconsistencies across platforms, and the 4B and 7B variants were reported to beat prior open models across six benchmarks on mobile, desktop, and web, including out-of-distribution settings. The open release is aimed at letting researchers train custom GUI agents without depending on closed commercial VLMs. Verify the license on the model card before commercial use.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 85; Trending score 76/100 from researcher assessment.