
Local LLM on the Apple Silicon Macs
I've been messing with on-device AI stuff for Apple Silicon Macs and put together LocalLM Lab to make experimentation easier.
It started as a simple prompt interface as just a way to poke at Apple's on-device foundation model and see how it responds. Since then I've added an OpenAI API-compatible endpoint (localhost or external IP), so you can point existing tools/scripts at it like any other local model server.
Right now it's built on Apple's Foundation Models framework. With the new LanguageModel protocol Apple introduced this year, the framework isn't locked to Apple's on-device model anymore. Any provider that ships a conforming Swift package (local MLX models, Claude, Gemini, etc) should be a plug-in. I'm looking at wiring that up so LocalLM Lab isn't limited to just the built-in model.
It's a free download from thisbrain.ai/locallm
Let me know how you think LocalLM Lab should evolve. MCP? Model switching in the UI? Something else?