Alibaba’s Tongyi Lab released MAI-UI, a new family of multi-modal, general-purpose GUI intelligent agents. The system enhances human-computer interaction. It achieves this by integrating tool usage, collaboration between devices and the cloud, and online reinforcement learning.
The lab claims MAI-UI achieves leading results in GUI navigation benchmarks. Testing showed superior performance on the MobileWorld and AndroidWorld benchmarks. MAI-UI reportedly outperformed competing models, including Gemini2.5Pro and Seed1.8.
This release marks progress in developing AI capable of efficiently handling complex operations on smart devices.