High-performance LLMs optimized for Apple Silicon Macs using MLX. Run state-of-the-art models locally on M1/M2/M3/M4 Macs with unified memory.