Nemotron 3 Nano Omni 30B-A3B

Overview

NVIDIA's omnimodal MoE: 30B total / 3B active, handling text, image, audio, and video in 256k context. Hybrid Mamba2-MoE architecture delivers 9x the throughput of competing open omni models. Released April 2026.

When to pick this model

• High-throughput omnimodal inference on NVIDIA hardware
• Single-GPU deployments needing text + image + audio + video
• Long-context multimodal analysis (256k)
• Production pipelines built on NVIDIA NIM
• English-only voice and video assistants

VRAM requirements by quantization

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Strengths

• Native omnimodal: text, image, audio, video
• 256k context window
• 9x throughput versus other open omni models
• Runs on a single GPU thanks to 3B active MoE
• First-class NVIDIA NIM pipeline

Limitations

• English-only
• Full multimodal requires llama.cpp or vLLM (Ollama is text-only)
• NVIDIA Open Model License is not Apache or MIT

Architecture & training

Architecture: Hybrid Mamba2-Transformer MoE · 30B total / 3B active · Conv3D + EVS · integrated vision/audio/video

Training: 354.6M samples · ~717B tokens across 1,395 datasets. English only. BF16, FP8, NVFP4 variants released.

Quick start

# HuggingFace : nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.