MODELS

NVIDIA: Nemotron 3 Nano Omni (free)

by NVIDIA

modellead-sourceopenrouter-modelsmultimodalnvidianemotronopen-modelhugging-faceopenrouternim

Overview

NVIDIA Nemotron 3 Nano Omni is an open multimodal model for video, audio, image, and text workflows.

Details

NVIDIA Nemotron 3 Nano Omni is described by OpenRouter as a 30B-A3B open multimodal model designed to act as a perception and context sub-agent in enterprise agent systems. NVIDIA’s NIM model card identifies it as a multimodal model for video, audio, image, and text workflows, with commercial use terms, release date, model versions, and Hugging Face weight downloads. The official NVIDIA Hugging Face model card lists the BF16 checkpoint, 31B total parameters, 256k max context, and deployment guidance.

When to Use

Use when an agent workflow needs a multimodal perception/context component across video, audio, image, and text inputs. Use when evaluating NVIDIA-hosted or NVIDIA-published open model weights for enterprise multimodal applications. Use via OpenRouter when specifically testing the listed free OpenRouter route before choosing a production deployment path.

Getting Started

Review the NVIDIA NIM model card for model versions, commercial use terms, and deployment context.
Inspect the official NVIDIA Hugging Face BF16 checkpoint page for weights, license, context length, and deployment guidance.
If using OpenRouter, test the nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free listing with a small multimodal prompt before broader evaluation.

Key Features

•Open multimodal model for video, audio, image, and text workflows.
•Designed to function as a perception and context sub-agent in enterprise agent systems.
•Official Hugging Face checkpoint lists 31B total parameters.
•Official Hugging Face model card lists a 256k maximum context.
•NVIDIA NIM model card lists commercial use terms, model versions, and Hugging Face weight downloads.

Capabilities

•multimodal-input
•text-input
•image-input
•video-input
•audio-input
•long-context
•agent-context-perception

Last updated May 31, 2026