NVIDIA introduced the open‑source Nemotron 3 Nano Omni multimodal AI model featuring a 30B‑A3B MoE architecture and a 256K context window.
The model delivers up to nine‑fold higher throughput than comparable open omni models, handling text, images, audio, video, documents, charts and GUIs.
Companies including Aible, Foxconn, Palantir and Infosys are adopting or evaluating the model; H Company’s CEO cites full‑HD visual reasoning.
It ships with open weights and is available on Hugging Face, OpenRouter, NVIDIA NIM microservice and partners, meeting regulatory and data‑localisation needs.