NVIDIA Nemotron 3 Nano Omni model now available on Amazon SageMaker JumpStart

This article announces the general availability of NVIDIA Nemotron 3 Nano Omni on Amazon SageMaker JumpStart, a multimodal model that processes video, audio, images, and text in a single inference pass.

• 30 billion total parameters with 3 billion active parameters using Mamba2 Transformer Hybrid MoE architecture
• Supports video (up to 2 minutes), audio (up to 1 hour), images, and text with 131K token context length
• Enables enterprise use cases: computer use agents, document intelligence, audio/video understanding
• Simplifies agentic workflows by eliminating fragmented model stacks and multiple inference passes
• Deploy via SageMaker Studio or Python SDK with one-click deployment
• Supports image understanding, video reasoning, and audio transcription inference modes
• Recommended parameters: thinking mode (temperature 0.6, max_tokens 20480) for complex reasoning

Nemotron 3 Nano Omni provides unified multimodal perception for enterprise agents, reducing latency and complexity compared to stitched-together separate models.