I'm curious why not. I am running a few different models on my mac studio. I'm using llama.cpp, and it performs amazingly fast for the $7k I spent.