NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4

Frontier accuracy, 5X greater speed, 30% lower cost.

A NVIDIA Nemotron 3 Ultra language-model release with 550B total parameters and 55B active parameters, using the NVFP4 checkpoint format. It is a hybrid Mamba-Transformer mixture-of-experts model release intended for long-context and agentic workloads.

Pricing

Model profile · Current snapshot

Input / 1M

$0.60

Output / 1M

$2.60

Blended / 1M

$1.10

Output TPS

164

TTFT (s)

1.26

Model Intelligence

Context window

1,000,000 tokens

Arena ranking

Benchmarkable

Yes

Model level

release

Intelligence Index

23.4

Coding Index

49.3

GPQA

0.87

HLE

0.28

SciCode

0.4

IFBench

0.81

LCR

0.79

TerminalBench Hard

0.36

TAU2

0.83

Recent stories

0 linked stories

No linked stories yet.