Carbon-X offers official GGUF quantized models including a 15.8 GB Q8_0 build and a 9.6 GB Q4_K_M build requiring llama.cpp build b11742 or newer