Prince Canuma, the developer of the MLX-VLM Python library, has released a new native macOS desktop application called Nativ that wraps MLX to run AI models locally on a Mac. The app aims to make MLX-powered vision and language models more accessible to users who prefer a graphical interface over command-line tools.
Features
- Chat interface: Nativ provides a conversation-style UI that lets users interact directly with loaded models.
- Localhost API server: the app runs a local HTTP API so other applications can access the models programmatically on the same machine.
The overall approach resembles LM Studio by combining a user-facing GUI with a local API for integrations.
Model detection from Hugging Face cache
Nativ can detect and pick up MLX-compatible models that are already present in the user’s Hugging Face cache directory. This makes it convenient for users who have previously downloaded models to the same machine.
Community mention
The project and early impressions were discussed on Hacker News, where users shared feedback on Nativ and its MLX integration.
Summary
Nativ is a native macOS application from Prince Canuma (author of MLX-VLM) that wraps MLX to provide both a chat interface and a localhost API for running models locally. Its ability to reuse models found in a Hugging Face cache is a practical convenience for existing users.
