GGML

GGML is a tensor library in C for machine learning inference on commodity hardware, originally written by Georgi Gerganov as the substrate underneath llama.cpp and whisper.cpp. It implements automatic differentiation, integer quantization, and hardware backends for CPU, CUDA, Metal, and Vulkan, and is the basis of the GGUF file format used across the local LLM ecosystem. Active development now mostly happens in those downstream projects.