Hugging Face has added support for running GGUF models efficiently in its Transformers library, Hugging Face said on 22 September 2026.
GGUF is the model format developed by the llama.cpp team and is widely used for running AI models locally. llama.cpp powers local tools including Ollama, LM Studio and Jan.
With the change, users can pick a GGUF checkpoint from the Hub, load it with the standard from_pretrained call and generate text on their own machine.
Quantised GGUF checkpoints are published on the Hub by the llama.cpp team under ggml-org and by publishers such as Unsloth, LM Studio Community and bartowski, in a range of sizes to fit different hardware.
Hugging Face said GGUF models have been downloaded millions of times and it wants to make them easier to use from Transformers.