Agent runtime infrastructure Multimodal Model serving Nemotron

NVIDIA Nemotron 3 Nano Omni

Audio, Video, Image, Text → Text

A multimodal NVIDIA Nemotron 3 Nano release described as a 30B-parameter, 3B-active model for text, image, audio, and video understanding with text output and a 256K-token context window.

Pricing

Model profile · Current snapshot

Input / 1M

$0.50

Output / 1M

$0.90

Blended / 1M

$0.60

Output TPS

232

TTFT (s)

0.5

Model Intelligence

Context window

262,144 tokens

Arena ranking

Benchmarkable

Yes

Model level

release

GPQA

0.47

HLE

0.05

IFBench

0.63

LCR

0.4

TerminalBench Hard

0.08

TAU2

0.45

Recent stories

1 linked story

releasePRIMARY2026-04-28

Nemotron 3 Nano Omni launches 30B-A3B multimodal model with 256K context

NVIDIA opened Nemotron 3 Nano Omni, a 30B-A3B model for text, image, audio, and video, with day-one serving support. That lets teams run one open model for perception-heavy agents instead of stitching separate components.