llama.cpp is a C/C++ implementation of LLM inference designed to run efficiently on consumer hardware, including machines with no dedicated GPU. It introduced the GGUF quantisation format now used across the local-model ecosystem, and a great many friendlier tools are wrappers around it. Working with it directly means compiling and managing models yourself, in exchange for maximum control over quantisation, threading and memory. It is the right layer if you want to understand what the convenience tools are doing.
Pricing: Free
