Record 011 · Runtimes · 2023
67%23.11.U
GGUF
llama.cpp community
“A single-file model format that replaced a pile of .bin shards and Python glue.”
GGUF succeeded GGML as the portable, metadata-rich container for quantized weights. Hugging Face became a GGUF storefront.
The interchange format the local ecosystem actually uses.
- Single-file quantized models
- Embedded metadata
- Cross-runtime portability
The file format of local LLMs.
