Nvidia launches Nemotron 3 Ultra, an open-source model that boosts AI speed by 5 times

Nvidia has released the open-source Nemotron 3 Ultra model, a 550-billion-parameter mixture-of-experts system designed for always-on AI agents. It's built by the Nemotron Alliance to boost AI capabilities in coding, research, and business tasks.

Compared to similar open-source models, Nemotron 3 Ultra runs 5 times faster and costs 30% less. It works with popular agent platforms like Hermes Agent and LangChain, making it easier for companies to deploy AI helpers. Nvidia also launched new safety and voice recognition models to expand its lineup.

The model is already being used by CrowdStrike for finding security bugs and Palantir for handling complex tasks on its AI platform. Nemotron 3 Ultra will be available on June 4 through Hugging Face and other services.