88Z AI, 21 August — Alibaba Tongyi has released Qwen-UI-Agent, a foundation model for GUI agents designed for mobile, desktop, and web scenarios. The model outperforms leading flagship models in various benchmarks and features strict safety confirmation mechanisms to ensure secure user operations.
At a glance
- Trained on over 100 real devices and 150+ apps, outperforming mainstream flagship models in multiple benchmarks.
- Features strict safety boundaries, rejecting illegal requests and actively confirming sensitive operations.
The update
Alibaba Tongyi has officially launched Qwen-UI-Agent, a foundation model for agents specifically designed for graphical user interface interactions. Its application scenarios extensively cover smartphones, personal computers, web browsing, and deep search tasks. To enhance the model's practical capabilities, the research team conducted specialized training based on over one hundred real physical devices and more than one hundred and fifty applications. In terms of performance evaluation, the model achieved leading results in multiple authoritative benchmarks such as MobileWorld, OSWorld, and WebArena, with comprehensive performance surpassing current mainstream flagship models like GPT, Claude, and Gemini. Furthermore, the model establishes strict safety boundaries, capable of automatically identifying and rejecting illegal instructions. Before executing high-risk sensitive operations such as payments or file deletion, the system actively requires secondary confirmation from the user, thereby effectively ensuring operational safety.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 千问大模型Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.