---
title: "Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents"
description: "NVIDIA Nemotron 3 Nano Omni is introduced as an omni-modal understanding model for document analysis, multiple-image reasoning, automatic speech recognition, long audio-video understanding, agentic computer use, and general reasoning. It extends the Nemotron multimodal line from vision-language processing to text, image, video, and audio, targeting long, dense inputs such as 100-plus-page documents and mixed audio-visual evidence. Its unified encoder-projector-decoder combines a Nemotron 3 Nano 30B-A3B hybrid Mamba-Transformer MoE backbone with C-RADIOv4-H and Parakeet-TDT-0.6B-v2 encoders; dynamic visual resolution, Conv3D tubelets, and Efficient Video Sampling support multimodal context compression. The post reports leading or top results across document, video, audio-video, voice, and ASR benchmarks, alongside up to 9x higher throughput and 2.9x single-stream reasoning speed on multimodal use cases, and provides BF16, FP8, and NVFP4 checkpoints."
---

# Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents

[Hugging Face](https://yomu.fyi/company/hugging-face) · Tuomas Rintamaki, Amala Sanjay Deshmukh, Nabin Mulepati, Collin McCarthy, Pritam Biswas, Arushi Goel, Alexandre Milesi, Danial Mohseni Taheri, Kateryna Chumachenko, Isabel Hulseman, Zhehuai Chen, Karan, Tao, Leili Tavabi, Ehsan Hosseini Asl · Apr 28, 2026

**Type:** Announcement

## Summary

NVIDIA Nemotron 3 Nano Omni is introduced as an omni-modal understanding model for document analysis, multiple-image reasoning, automatic speech recognition, long audio-video understanding, agentic computer use, and general reasoning. It extends the Nemotron multimodal line from vision-language processing to text, image, video, and audio, targeting long, dense inputs such as 100-plus-page documents and mixed audio-visual evidence. Its unified encoder-projector-decoder combines a Nemotron 3 Nano 30B-A3B hybrid Mamba-Transformer MoE backbone with C-RADIOv4-H and Parakeet-TDT-0.6B-v2 encoders; dynamic visual resolution, Conv3D tubelets, and Efficient Video Sampling support multimodal context compression. The post reports leading or top results across document, video, audio-video, voice, and ASR benchmarks, alongside up to 9x higher throughput and 2.9x single-stream reasoning speed on multimodal use cases, and provides BF16, FP8, and NVFP4 checkpoints.

## Context

The model is positioned for long, complex multimodal workloads spanning high-value documents, diverse and lengthy audio, mixed audio-video evidence, GUI environments, and reasoning across text, images, tables, and other structured or semi-structured inputs. It extends an earlier Nemotron vision-language system with native audio and broader video capabilities.

## Approach / What changed

Nemotron 3 Nano Omni uses a unified encoder-projector-decoder architecture with a Nemotron 3 Nano 30B-A3B hybrid Mamba-Transformer MoE language backbone, C-RADIOv4-H vision encoder, and Parakeet-TDT-0.6B-v2 audio encoder. Its design adds dynamic-resolution vision, Conv3D temporal compression, Efficient Video Sampling, and native audio input. Training uses staged multimodal alignment and context extension followed by preference optimization and multimodal reinforcement learning.

## Takeaways

- The backbone interleaves 23 Mamba selective state-space layers, 23 MoE layers with 128 experts, top-6 routing and a shared expert, and 6 grouped-query attention layers.
- Each image can use between 1,024 and 13,312 visual patches of size 16 x 16, allowing variable resolution for documents, charts, slides, screenshots, and GUI layouts.
- Conv3D fuses every pair of consecutive video frames into one tubelet, while EVS removes redundant static tokens after the vision encoder to reduce latency and improve throughput.

**Tags:** [LLMs](https://yomu.fyi/topic/llm), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance)

- Source: [Hugging Face](https://huggingface.co/blog/nvidia/nemotron-3-nano-omni-multimodal-intelligence)
- Source URL: https://huggingface.co/blog/nvidia/nemotron-3-nano-omni-multimodal-intelligence
- Ingested by Yomu: 2026-08-29T01:48:26.795Z

[Read original post](https://huggingface.co/blog/nvidia/nemotron-3-nano-omni-multimodal-intelligence)