Hugging Face Transformers runs GGUF models at llama.cpp speed
Developers who serve GGUF-quantized open-source models can now run them directly through Transformers with llama.cpp-level speed, removing a common reason to keep two separate inference stacks.
