MLX

MLX is an open-source machine learning framework Apple ML Research released in late 2023, designed from the ground up for Apple Silicon (the M-series chips). It exposes a NumPy-like Python API but exploits the unified-memory architecture of M chips and the Neural Engine (NPU) efficiently — making local training and inference of modern LLMs unusually fast on Apple hardware. With Hugging Face transformers integration and libraries like mlx-lm, running Llama 3, Mistral or Qwen on a Mac has reached a 'nearly zero extra setup' level of convenience. MLX is one of the most concrete developer-facing signals of where the Apple Intelligence ecosystem is heading.