NVIDIA Launches Nemotron 3 Nano Omni: Open-Source Omni-Modal Model Supporting Millions of Tokens

NVIDIA has released Nemotron 3 Nano Omni, an open-source omni-modal model that continues the "Nano" positioning, emphasizing high cost-performance and inference efficiency. The model involves approximately 30 billion total parameters and supports ultra-long context windows of up to one million tokens. The model adopts a 30B-A3B Mixture-of-Experts (MoE) architecture and integrates Mamba layers with Transformer layers at the architectural level. According to official data, this hybrid design improves memory and computational efficiency by up to 4 times. When benchmarked under a fixed user interaction latency threshold, the model achieves approximately 9.2× higher effective system capacity for video reasoning tasks compared to other open omni-modal models, and approximately 7.4× higher effective system capacity for multi-document reasoning tasks. Companies in AI and software sectors such as Foxconn and Palantir have already adopted the new model, while Dell and Oracle are currently evaluating it.