UI-TARS
Open vision-language-action model that operates computers by looking at the screen
About
UI-TARS is ByteDance's open vision-language-action model for controlling computers through screenshots, released under Apache 2.0 with the paper published in January 2025. Instead of reading DOM trees or APIs, it looks at the screen the way a person does, reasons about the next step, and clicks and types accordingly. A UI-TARS Desktop app and SDK wrap the model for cross-platform use, though visual grounding can drift and complex multi-step tasks still need supervision.
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: Paper published January 2025; ByteDance + Tsinghua collaboration; Apache 2.0 confirmed by multiple sources; UI-TARS Desktop app v0.1.0 (2025-04-17) and SDK verified from repo news; 32k+ GitHub stars reported by third parties; No verbatim official repo URL surfaced in search results, so officialUrl left empty rather than guessed; releasedAt left null: paper month confirmed (Jan 2025), exact day not pinned. License: Apache-2.0 confirmed across sources..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Paper published January 2025; ByteDance + Tsinghua collaboration; Apache 2.0 confirmed by multiple sources
- · UI-TARS Desktop app v0.1.0 (2025-04-17) and SDK verified from repo news; 32k+ GitHub stars reported by third parties
- · No verbatim official repo URL surfaced in search results, so officialUrl left empty rather than guessed
- · releasedAt left null: paper month confirmed (Jan 2025), exact day not pinned