Nativ
The developer whose libraries quietly power LM Studio's Mac engine has released his own fully open app that does the same job, and made the whole thing MIT.
Prince Canuma maintains MLX-VLM and MLX-LM, the open libraries that let Apple-silicon Macs run vision and language models fast. LM Studio, the polished app most people reach for to do exactly that, credited those libraries by name when it added Mac support. This week Canuma shipped Nativ, his own native Mac app for local inference, and released it under a permissive MIT licence including the Swift app itself. The person who built the engine the popular closed app runs on now has an open app of his own.
For a reader with an M-series Mac, the appeal is plain: download one free app, no account or subscription, and it scans your existing model folder and starts serving. Simon Willison confirmed it found MLX models already sitting in his Hugging Face cache without a re-download. Beyond the chat window it opens a local server on your machine that speaks both OpenAI's and Anthropic's API formats, so tools like Claude Code or Codex can point at it and run entirely offline against open models from Google, Cohere, and Liquid AI.
The category is crowded. LM Studio, Ollama, and Jan already cover most of this ground, and Nativ is a version 0.0.1 with a few hundred stars. What is worth noting is the provenance rather than the feature list: the maintainer closing the loop on his own stack, and shipping the app layer open too. Judge it on that and on the code, not on a day-two star count.
The lenses
The facts
Concepts
How this connects
Tap a node to open it