[LoganDark]

llama.cpp uses GGML which uses Metal directly.