MLX Studio is a native macOS application for running AI models locally on Apple Silicon computers. Powered by Apple's MLX framework and the vMLX inference engine, it provides GPU-accelerated processing without complex setup. Users can download models directly from Hugging Face and interact through a conversational interface. It supports multimodal image understanding, reasoning models, and structured tool calling. Additional capabilities include AI image generation, text-to-speech synthesis, and Whisper-based transcription. An OpenAI-compatible API server connects local models to external applications. Developer tools support model inspection, conversion, quantization, and diagnostics. Conversations can be saved and organized locally.
Features
- Local GPU-accelerated AI inference optimized for Apple Silicon Mac computers
- Hugging Face model downloading with multimodal chat and reasoning support
- AI image generation with MLX-powered diffusion models and text-to-speech synthesis
- Whisper speech transcription and structured function calling for agent workflows
- OpenAI-compatible local API server with interactive documentation and testing tools
- Model conversion, quantization, diagnostics, and persistent conversation management