llama.cpp

A lightweight, high-performance C/C++ implementation for running LLaMA‑style large language models locally, supporting CPU and GPU inference with minimal dependencies.