Apple Silicon and macOS VMs: 11–16× Faster LLM Inference with Llama.cpp

Guide walks through GPU passthrough for macOS VMs on Apple Silicon, reporting 11–16× faster LLM inference with llama.cpp.