​MLX is an array framework for efficient and flexible machine learning on Apple silicon.

MLX can be run on any Apple platform that supports Metal:
Some key features of MLX are:
  • A familiar API based on NumPy
  • Designed for unified memory
  • Composable function transformations
  • Multi-device support (CPU or GPU)

Get started with MLX on Apple silicon.

  • MLX LM - a Python package for generating text and fine-tuning language models on Apple silicon.
  • MLX Whisper - a Python package for speech transcription using OpenAI’s Whisper models.
  • MLX Examples - standalone examples in the MLX framework including image generation, speech and music generation, language model training and many more.
  • MLX Swift Examples - examples in MLX Swift including LLM and VLM text generation, low-rank fine-tuning, and image generation with SDXL.

Check out the growing community.

In addition to the Apple-released examples, there are a growing number of community-built projects that show how the framework can be used to bring a wide range of models to Apple silicon.