Funny, I'm running local LLMs in a modest iMac M4, they are slow indeed but not utterly slow. A dedicated system should be way faster I'd guess.