Blaizzy/mlx-vlm
★ 5,809 stars
Covered in week 2026-W21 · the 2026-05 report
MLX-VLM is a package for inference and fine-tuning of Vision Language Models (VLMs) on your Mac using MLX.
What mlx-vlm is
The easiest way to get started is to install the mlx-vlm package using pip
From the repository README · read it on GitHub →
Transcript
GPT-4 Vision costs $10 per million tokens for image analysis. MLX-VLM does it free, right on your Mac. This open source tool lets you run powerful vision language models like LLaVA, Pixtral, and Qwen locally on Apple Silicon with zero cloud costs. Over 4,700 developers have already starred it. You can analyze images, describe photos, and even fine-tune models, all without sending a single file to the cloud. Star it if this saved you money.