Local LLMs thrive on Apple's hardware, and a huge part of it is thanks to MLX.
Programming model moves from managing thousands of low-level threads to working with high-level ‘tiles of data’ ...