Blaizzy/mlx-vlm: Vision Language Model Inference and Fine-Tuning on Apple Silicon — 3K Stars, +382/Day
GitHub·medium signal
MLX-VLM enables running and fine-tuning vision language models and omni models locally on Apple Silicon Macs using the MLX framework. Supports image, audio, and multi-modal inputs with CLI, Python, and FastAPI server deployment options. Recent additions include activation quantization for CUDA, thinking budget controls, and expanded model support. Growing at +382 stars/day as local/edge AI inference demand surges. 506 commits indicate mature, actively maintained codebase.