NVIDIA introduces NVFP4

NVFP4 introduced — an FP8 scale for every 16-value micro-block

Why it matters

A precise standardization of 4-bit precision

Source — Primary sources first; secondary reports are labelled “Press”

Officialdeveloper.nvidia.comhttps://developer.nvidia.com/blog/introducing-nvfp4-for-efficient-and-accurate-low-precision-inference/

Added in Week 35, 2026 · See that week

Same week

The term context engineering becomes establishedOfficialGoogle officially releases Gemma 3nOfficialPartial summary judgment in Kadrey v. MetaOfficialA2A project launches at the Linux Foundation, backed by 100+ companiesOfficial

Nearby on the same axis

PyTorch Foundation takes in vLLM as a hosted projectOfficialSGLang reproduces large-scale EP on 96 H100sOfficialGemma 3 QAT int4OfficialDeepSeek-V3.2-ExpOfficial