ollama run llama3:70b It seems to run well and feels like GPT-4, with the added benefit that there’s no “limit replies as much as possible” system restriction. However it pukes itself pretty quickly and starts emitting garbage.

→ View original post on X — @steipete,