Z-Ai/Autoglm Phone 9b is a compact vision language model aimed squarely at smartphone automation. It inherits the GLM-4V-9B backbone and is tuned to read phone screens as images, identify interface elements, and emit the precise taps, swipes, or text entries needed to carry out a user request. To process visuals it pairs the language weights with a multimodal projector file, letting it convert rendered UI state into actionable instructions rather than only describing what it sees.

For developers building agent-style mobile workflows, the appeal is the tight coupling between visual perception and tool-ready output. The 9B-parameter footprint keeps inference light enough for device-adjacent or edge-hosted scenarios, while the screen-aware training makes it well suited to GUI navigation tasks where traditional text-only assistants fall short. Community quantizations circulated as GGUF builds further indicate flexibility for self-hosted deployments and experimentation outside the hosted API.