Zhipu AI has open-sourced AutoGLM, an AI agent model described as the first to demonstrate stable on-device “phone use” capabilities, Blue Whale Finance reported. The model can perform multi-step tasks such as food delivery orders and flight bookings by interpreting on-screen content and simulating human-like taps, swipes, and text input. AutoGLM currently supports core workflows across more than 50 high-frequency Chinese apps, including WeChat, Taobao, Douyin, and Meituan.
Zhipu AI said hardware makers, smartphone vendors, and developers can use the release to reproduce a fully functional phone-control assistant on their own systems. The open-source package includes trained models, a phone-use framework and toolchain, runnable demos, Android adaptation layers, and documentation. The company noted that the project supports both local and cloud deployment to maintain user control over data and privacy. [TechNode reporting]
Related
