Hacker News new | past | comments | ask | show | jobs | submit
The M7 will the much better for local LLM inference. If you can afford the memory.
Oh interesting, any more info on "how"?
loading story #50263312