NVFP4

precision

Also known as: NVIDIA FP4

Definition

NVIDIA's own 4-bit floating-point format for Blackwell, using finer-grained scaling blocks than standard FP4 to hold accuracy at very low precision. Quoted separately from FP4 on some parts because the two are not numerically identical.

GPUs featuring NVFP4

NVIDIA Rubin

Rubin · 2026 · 288 GB

NVIDIA Vera Rubin

Rubin · 2026 · 576 GB

Browse All GPUs →

Related Terms

FP4

A 4-bit floating-point format introduced with NVIDIA Blackwell, halving the memo...

FP8

8-bit floating-point precision format introduced in Hopper architecture. Enables...

Transformer Engine

Hardware and software technology in Hopper architecture designed to accelerate T...

More precision Terms

TF32

TensorFloat-32 - A 19-bit precision format introduced with NVIDIA Ampere archite...

FP16

16-bit floating-point precision format (half precision). Commonly used in AI tra...

BF16

Brain Float 16 - A 16-bit floating-point format developed by Google with the sam...

Explore More Terms

Browse the full glossary or dive into architecture guides

Back to Glossary Architecture Guides