Llama.cpp 30B runs with only 6GB of RAM now by from Hacker News on 2023-03-31 20:37 (#6ACMA) Comments