Nvidia introduces Nemotron 3 open models for multi agent AI

Nvidia has introduced the Nemotron 3 family of open models, data and libraries, which are designed to power transparent, efficient and specialized agentic artificial intelligence development across a wide range of industries. The company positions Nemotron 3 as a foundational toolkit that can be used to build and deploy advanced artificial intelligence agents that can work together on complex tasks while remaining understandable and controllable for developers and enterprises.

The Nemotron 3 models, available in Nano, Super and Ultra sizes, feature a hybrid latent mixture-of-experts architecture that Nvidia describes as a breakthrough design for multi agent systems at scale. This hybrid latent mixture-of-experts approach is intended to help developers construct reliable agent workflows by routing different parts of a problem to specialized expert components within the model family, while keeping the overall system efficient enough for real world deployment.

Nvidia frames Nemotron 3 as a response to a broader industry shift from single model chatbots to collaborative multi agent artificial intelligence systems, where multiple agents must coordinate, share context and handle different segments of a workflow. Developers in these environments face challenges such as communication overhead between agents, context drift over time and high inference costs, as well as a need for transparency so they can trust automated decision making. Nemotron 3 is presented as directly addressing these issues by combining open models, accessible data and supporting libraries to deliver both performance and openness for building specialized agentic artificial intelligence applications.