Comment by nicce

> Btw, you can also run Mistral locally within the Docker model runner on a Mac.

Efficiently? I thought macOS does not have API so that Docker could use GPU.