[falava]

Performance of llama.cpp on Apple Silicon M-series:

https://github.com/ggerganov/llama.cpp/discussions/4167