prima.cpp

github.com ↗

A distributed implementation of llama.cpp that lets you run 70B-level LLMs on your everyday devices.