llama.cpp

20a04b22 - ggml-cpu: use UE4M3 LUT in ARM NVFP4 dot product (#25331)