NVIDIA released Nemotron 3 Nano Omni on April 28, a 30B-A3B hybrid mixture-of-experts open model that processes text, images, audio, and video natively in a single system with a 256K-token context window and 1920×1080 native image resolution. Unlike pipeline architectures that chain separate perception models, the unified design maintains context across modalities and achieves up to 9x higher throughput than comparable open omni models on video and document workloads, while running on 25 GB of RAM. The model is available through Hugging Face, OpenRouter, NVIDIA NIM, and 25+ partner platforms, with early enterprise adopters including Foxconn, Palantir, and Docusign.