How to Use Apple's fm CLI to Run On-Device LLMs From Python

Apple's `fm` command-line interface is a native macOS utility for running on-device foundation models locally on Apple Silicon, offering zero-cost, private inference. Practical tests of its text reasoning and multimodal vision capabilities revealed significant limitations, including factual inaccuracies and image misclassifications. These results suggest the tool is best suited for local prototyping and development workflows rather than as a production-ready knowledge base. Developers can integrate the model into applications using the `fm serve` command, which exposes an OpenAI-compatible endpoint for easy use in Python and other environments.