For specifically unified memory reasons, Apple machines are much, much better at running local LLMs.