31B total / 3B active (30B-A3B class MoE)
256K tokens
NVIDIA Open Model Agreement
Official open-weight multimodal Nemotron release for enterprise-grade omni-understanding across video, audio, image, and text inputs with text output. NVIDIA positions Nemotron 3 Nano Omni as a native multimodal reasoning model for document intelligence, meeting and media analysis, OCR-heavy workflows, transcription, and GUI automation for agentic applications. The model card describes a Mamba2-Transformer hybrid Mixture-of-Experts architecture built on the Nemotron 3 Nano 30B-A3B backbone, extended with a CRADIO v4-H vision encoder and Parakeet speech encoder. Release materials emphasize integrated video plus speech comprehension, OCR reasoning, GUI understanding, JSON output, tool calling, word-level timestamps for transcription, and configurable reasoning output. The Hugging Face card lists support for up to 2-minute video inputs, audio files up to 1 hour, and maximum context length up to 256K tokens, with BF16, FP8, and NVFP4 public weight releases published together. NVIDIA also highlights enterprise deployment support through vLLM, TensorRT-LLM, SGLang, llama.cpp, and Ollama-oriented workflows.