2 ms·

I use lower parameter LLM's on M1's with very little RAM as well. but 70B+ would be useful for me too