Nvidia Releases Multi-modal AI Model Nemotron3NanoOmni with 9x Higher Throughput Than Peers

Nvidia recently launched the open multi-modal AI model Nemotron3NanoOmni, integrating advanced reasoning capabilities for video, audio, images, and text into one system. Using a 30B-A3B mixed expert architecture, its throughput is 9 times higher than similar models, providing an efficient path for enterprises and developers to build intelligent agents.

Nemotron3NanoOmni ranks first in six rankings for complex document intelligence, video, and audio understanding, with high accuracy and low cost. It can also work with proprietary cloud models or other Nvidia Nemotron open models, offering great flexibility. Additionally, the Nemotron3 series has been downloaded over 50 million times in the past year, showing strong market demand.

Gautier Cloix, CEO of HCompany, said that using this model, their agents can quickly interpret full HD screen recordings—something impossible with previous technologies. The launch of this model will drive the development of multi-modal AI, helping enterprises improve user experience and competitive advantages.