Qwen-UI-Agent
Alibaba · 27B main / 35B-A3B / 4B tiers
The open-weight leader in GUI agents
A GUI-agent family released 20 August 2026 in 27B, 35B-A3B and 4B sizes under Apache-2.0: MobileWorld-Real 92.2 and AndroidDaily 97.5 give it agents 98, the new single-dimension leader here. The trade-offs are explicit: generation 50, video 60 and documents 80 — this is not a general multimodal model but an executor for phone and desktop control.
Last updated Oct 21, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- Qwen-UI-Agent79.5
Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.
Public benchmarks
as reported, not converted| Benchmark / metric | Value | Note |
|---|---|---|
| MobileWorld-Real | 92.2 | Official |
| AndroidDaily | 97.5 | Official |
| 许可证 | Apache-2.0 | Three open-weight sizes |
Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.
Scenario mapping
five stress tests, one scale| Scenario | Score | Observation |
|---|---|---|
| Mobile GUI control | 98 · Single-dimension leader here | |
| Chinese screenshot OCR | 80 · Reads UI text; weak on complex tables | |
| Chart data extraction | 70 · Simple charts only | |
| 30-min video QA | 60 · Not a video model | |
| Chinese cultural context | 88 · Strong in task context |
Strengths and limitations
Strengths
- Agents 98: the new single-dimension leader here, above Kimi K3 at 96.
- Three sizes under Apache-2.0, with the 4B edge and 27B cloud tiers sharing one action space.
- Official MobileWorld-Real and AndroidDaily real-device benchmarks make cross-checking easier.
Limitations
- Generation 50 and video 60: it operates, it does not understand content.
- Documents 80 and charts 70: complex tables and charts in screenshots are easily misread.
- Long real-device tasks still drift; there is a gap between vendor benchmarks and live apps.
Price and real cost
List price
—N/A
Open weights; self-hosted, no API fee. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥0.0 (assuming input at 25% of output price).
Use your volumeAbout Qwen-UI-Agent
Is Qwen-UI-Agent open source?
Licence: Open weights (Apache-2.0). Weights are available, but check the exact terms and filing requirements before commercial use.
How much does Qwen-UI-Agent cost?
Open weights; self-hosted, no API fee. It ranks #38 with a composite of 79.5. Real bills depend on resolution, tiling, caching and volume discounts.
What is Qwen-UI-Agent best for?
Mobile app automation testing; GUI agents and RPA; Accessibility and assisted control。
Why does Qwen-UI-Agent score 79.5?
The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is GUI agents & tool use; its weakest is Generation & editing.
Personal ranking · Not an official score
Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.