Nativ brings local AI model running to Mac with desktop app

On July 21, 2026, Prince Canuma, the developer of the MLX-VLM Python library (a specialized tool for running vision-language models using MLX on a Mac), released Nativ, a new macOS desktop application. Nativ wraps the MLX framework—which is optimized for running AI models efficiently on Apple hardware—into a full graphical application that runs locally on a user's Mac.

The application offers two main modes of access: a direct chat interface for interactive conversations with AI models, and a localhost API server that allows developers to query models programmatically from their own applications. This dual approach mirrors the functionality of LM Studio, an existing tool in the same category.

One noteworthy feature is that Nativ automatically detects and loads MLX models the user has already cached in their Hugging Face cache directory. This design choice eliminates setup friction for users who have experimented with local AI models before, as the app recognizes existing downloads rather than forcing a fresh installation.

Prince Canuma, the developer behind the MLX-VLM Python library (a tool for running vision-language models on Apple hardware), has extended his work into a full consumer-facing application. Nativ wraps the MLX framework in a native macOS interface, mirroring the design of existing local-AI tools like LM Studio: it offers both an interactive chat mode and a localhost API server, giving users flexibility to integrate models into their own applications. The app's ability to auto-discover models from the Hugging Face cache directory—a standard cache location for users experimenting with open-source AI—lowers the friction for adoption among Mac users who already have models on disk.