The llama.cpp repository has merged support for NVFP4 quantization for LFM3 GGUF models. This enables higher-performance inference using NVIDIA's 4-bit floating point format.
HOW THIS AFFECTS YOU
●
builderYou can now achieve better inference performance and lower memory footprints on supported NVIDIA hardware.
●
researcherThis enables more efficient testing of new architectures using optimized quantization formats.