What is mlxui?

MLXUI is a native macOS app that lets you run local AI models without touching a terminal or writing Python. It is built on Apple's MLX framework, which is designed specifically for Apple Silicon chips like the M1 and later. The app gives you a visual way to browse, download, and run AI models that live entirely on your Mac, with no data sent to external servers and no per-token pricing or API keys required. The app ships with a curated catalog of around 28 models across seven categories: chat and text, vision, OCR, speech-to-text, text-to-speech, and embeddings. These include familiar names like Llama, Qwen, Gemma, and Whisper. The catalog tracks roughly 435 models from the MLX community on Hugging Face, and new models are added with each release. You can browse by category, filter by what fits your Mac's RAM and chip bandwidth, and install with a single click that downloads the model directly from Hugging Face. Each model type gets its own dedicated interface rather than a generic wrapper. Chat models open a chat window. Speech-to-text models give you a microphone and a live transcript. Text-to-speech models offer a voice picker and an audio player. Vision models let you upload an image and ask questions about it. OCR models provide a drop area for images and show extracted text. A pipeline runner lets you chain models together, for example transcribing audio, summarizing the text, and then speaking the summary back. The codebase uses a module system where each model type lives in its own isolated folder. A contributor adding image generation support does not need to understand the chat pipeline, and updating one speech module never touches another. The app is built in Swift and requires macOS 14.0 or later, an Apple Silicon Mac, and Xcode 15.0 or later to build from source. The full README is longer than what was shown.