MLX
More info
- Creator
- Apple
- Released
- Owner
- Apple
- Type
- ML framework for Apple Silicon
- License
- MIT
- Website
- ml-explore.github.io
- Source
- github.com/ml-explore/mlx
Apple's open-source ML framework optimised for Apple Silicon — exploits unified memory (CPU and GPU share RAM) to avoid data copies.
Released December 2023. The API is deliberately PyTorch-like with NumPy syntax; lazy evaluation lets operations fuse before they execute on the GPU. Built-in quantisation and support for Mistral, Llama, Phi, Stable Diffusion.
Sweet spot: local inference and small-scale fine-tuning on the M-series. An M2 Max with 64 GB unified memory can run Llama 3.1 70B Q4 locally — something that otherwise needs an A100. Not a competitor to big training stacks (JAX/PyTorch) but a strong contender for "consumer AI".