A Sovereign Signal project
01EntryItem
Archive
Runs LLMs larger than system RAM in pure C on CPU or GPU, claiming 37 tok/s on an 80B model and faster reads than llama.cpp on a 32 GB laptop.