Nativ: Prince Canuma Ships an MLX-Backed Mac App That Turns Cached Hugging Face Models Into a Localhost API Server

simonwillison.net·medium signal

Willison covered Nativ on July 21 — a macOS desktop app wrapping MLX that provides both a chat interface and a localhost API server for running vision-LLMs locally, and which auto-detected models already cached in his Hugging Face directory. The auto-detection detail is the actual product insight: the friction in local inference was never the runtime, it was rediscovering which weights you already downloaded. Paired with Kimi K3's imminent open-weight drop, the local-first stack for Apple Silicon builders got materially more usable this week.