much faster on M1 Mac with llama.cpp server

[scalar27]

FYI -- At least on my M1 Mac, running server (from llama.cpp) with Llama 3 8B is much faster for browsing the Dead Internet. I also tried ollama and LM studio with the same model. I guess I shouldn't be too surprised, but thought it might be useful for others.