Zhipu announced today the official open-source release of its core AI Agent model AutoGLM. This intelligent agent, which has the "Phone Use" capability, can stably complete complex mobile phone operation processes that involve dozens of steps, such as food delivery orders and flight bookings. The open source means that any hardware manufacturer, phone manufacturer, or developer can replicate an AI assistant in their own system that can "understand the screen" and perform operations like clicking, typing, and swiping, just like a human.

Using a phone, mobile internet

AutoGLM already supports more than 50 frequently used Chinese applications, including WeChat, Taobao, and Douyin, for their core scenarios. Its demonstration effect is very similar to the previously industry-recognized "Doubao Phone": users do not need to operate manually; they just need to let the AI observe the screen content, and it can automatically perform tasks, complete multiple-step operations continuously, and reach the result. Compared with other intelligent agent solutions, AutoGLM's advantage lies in its stability and ability to handle complex processes, making it especially suitable for executing long-chain tasks in real mobile environments.

This open-source move will significantly reduce the technical barriers for AI phones, making "AI can operate your phone" no longer a proprietary technology of top manufacturers but an ability that the entire industry can build together. It will promote the AI phone ecosystem to move from closed to open, and encourage more devices to have a system-level intelligent agent experience. AutoGLM also supports both local and cloud deployment, giving manufacturers maximum control over data and privacy, which also means users may be able to obtain the same level of intelligent experience without uploading private data in the future.

For smartphone manufacturers trying to build the next generation of system-level AI, the open-source release of AutoGLM is undoubtedly a strategic complement; for developers, it provides a complete intelligent agent capability foundation that is reproducible, modifiable, and scalable. As more manufacturers join the ecosystem, these AI agents with real interaction capabilities may accelerate to become standard features of future phones.