Open Models Almanac
Methodsinference & servinginference quantization

specific method · filed under inference & serving

NVFP4 quantization

NVFP4 stores each value in four bits (E2M1, fifteen distinct values from -6 to 6) and gives every block of 16 values one FP8 scale, plus one 32-bit scale for the tensor, which comes to 4.5 bits per value. The finer block and the non-power-of-two scale take twice MXFP4's scale overhead and put the scale closer to each block's actual maximum.

Written for the explainer Quantization for serving: which bits to remove, and what each removal saves.

Also called NVFP4 4-bit floating-point quantization, NVFP4 checkpoint, NVFP4 (NVIDIA 4-bit floating point), NVFP4, NVFP4 quantization checkpoint, NVFP4 quantized checkpoint serving.

sources
4
models
4
labs adopt it
3
strongest
used

Documented in

Evidence

4 quotes from 4 sources, one card per model, each opening with its strongest.

How sources treat it
optional 2used 2

Further reading

Picked by hand, not extracted: where to read more, not evidence for anything on this page.

↑