Rating certificate · AW/26-09/90326 Sept 2026
Not enough evidence
The Register / Browser & computer use / UI-TARS Desktop
Register entry · Browser & computer use · updated 26 Sept 2026
UI-TARS Desktop
UI-TARS-desktop (also known as Agent TARS) is ByteDance's Apache-2.0 open-source multimodal desktop agent stack that enables autonomous control of web browsers and operating systems. Built on ByteDance's vision-language models (UI-TARS 2B, 7B, 72B), it perceives raw screenshots directly to plan and execute mouse and keyboard actions without requiring DOM trees or accessibility labels. It can be run fully locally with self-hosted models or connected to cloud model endpoints.
Unverified description.The description, features and status on this entry come from a research batch that failed our source checks and have not been re-verified. Its pricing and dated changes are withheld; check the maker’s site.
What reviewers and users praise
- Not enough evidence yet.
What they criticise
- Not enough evidence yet.
Experts vs the crowdneeds both scores
Experts—
Crowd—
| Price | See the maker's site |
|---|---|
| Use it via | Desktop app, CLI, API |
| Runs on | macos, windows, linux |
| Licence | Apache-2.0 |
| Updated | Evidence collected 26 Sept 2026 |
How the rating was worked out
Weights for Browser & computer use: adoption 35%, experts 40%, crowd 25%. Fewer than two pillars have enough evidence. The method.
| Adoption & momentumweight 35% | 55 | GitHub stars 39k → 79 · 0 releases in 90 days → 0reach × 0.7 + momentum × 0.3 |
|---|---|---|
| Expertsindependent reviews · weight 40% | — | Only 1 independent review (3 needed) |
| Crowdcommunity sample · weight 25% | — | Only 3 community items sampled (6 needed) |
| Rating | — | Not rated: fewer than two pillars have enough evidence. |
Evidence checks. Part of this entry came from a research batch that failed verification: only items confirmed at their source are used, and its trust grades and pros and cons are withheld. 4 items removed: could not be verified at the source (4). How evidence is checked.
Trust & safety
Seven dimensions graded A–D from documented facts. Overall: not enough data.
n/e
Permission model
Not established.
n/e
Data access scope
Not established.
n/e
Data storage
Not established.
n/e
Data retention & training
Not established.
n/e
Incident history (24 months)
Not established.
n/e
Compliance
Not established.
n/e
Transparency
Not established.
The experts
1 named independent reviews found; dated reviews from the last 12 months are scored.
- positive
Stagehand and Browser Use are strong browser-first starting points; CUA, Agent S3, and UI-TARS target broader computer control
Alyona Vert. (Turing Post) · 2 Aug 2026 · turingpost.com
The crowd
3 coded posts from hackernews.
2 positive1 mixed0 negative
- positive
I've had lots of success with generating coordinates and answering questions using the UI-TARS model
Vision agents thread · 5 May 2026 · news.ycombinator.com - positive
I’d also checkout midscene, you can set the model and UI-TARS works but you can also use qwen vision models and it works.
Vision agents thread · 6 May 2026 · news.ycombinator.com - mixed
Looking at the table, I will admit that I don't get most of the use cases ( maybe with exception of comparison shopping
Computer use discussion · 27 Nov 2025 · news.ycombinator.com
Ranked alongside
Other ranked entries in Browser & computer use.
Browser UseBrowser Use (browser-use.com)66
StagehandBrowserbase65
Sources
Every figure above traces to one of these, observed up to 26 Sept 2026.