llama.cpp merges PR #19769, adding NVFP4
Merged PR #19769 — introduced GGML_TYPE_NVFP4. FP4 E2M1 weights + UE4M3 block scales, loading NVIDIA ModelOpt outputs directly
Why it matters
GGUF absorbed a hardware-native FP4 format beyond its own K-quants
Source — Primary sources first; secondary reports are labelled “Press”
Officialgithub.comhttps://github.com/ggml-org/llama.cpp/pull/19769Added in Week 35, 2026 · See that week