Here's a recipe for running the Qwen2-VL vision LLM models on Apple Silicon using Python and the mlx-vlm library, via a uv shell one-liner

Full details on my blog: https://simonwillison.net/2024/Sep/29/mlx-vlm/ - and here's the full output from that example prompt https://gist.github.com/simonw/9e02d425cacb902260ec1307e0671e17