Nvidia Introduces Llama Nemotron Open-Source LLMs to Build and Deploy AI Agents at CES 2025

Nvidia introduced the Llama Nemotron household of open massive language fashions (LLMs) on Monday. The firm stated that with the rise of synthetic intelligence (AI) brokers, new and extra refined AI fashions had been required to deal with the workflow of agentic AI. Highlighting the necessity for extra energy and better effectivity, the tech large acknowledged that the Nemotron household fashions can create and deploy AI brokers throughout numerous purposes. The firm claimed that the AI fashions will likely be obtainable for enterprises by way of the Nvidia NIM microservice.

Nvidia Introduces Nemotron Family of AI Models

In a blog post, the tech large introduced its new sequence of open-source LLMs dubbed Nemotron. The sequence additionally comprises Cosmos Nemotron imaginative and prescient language fashions (VLMs), and these can be utilized to construct AI brokers that analyse and reply to pictures and movies. Nvidia stated the vision-focused brokers could be deployed in autonomous machines, hospitals, shops and warehouses, in addition to sports activities occasions, films, and information.

Built with Meta’s Llama basis fashions, the Nvidia Llama Nemotron fashions are stated to be optimised to construct and develop AI brokers. While the corporate didn’t reveal the structure and technical particulars, it claimed that these fashions are educated utilizing “newest strategies and high-quality datasets”. The fashions can be utilized to coach agentic capabilities equivalent to instruction following, chat, perform calling, coding and arithmetic, and extra. Nemotron can be stated to optimise the AI brokers’ dimension to make it straightforward to deploy.

Nvidia acknowledged that SAP, ServiceNow, and different AI agent platform suppliers will likely be among the many first to make use of the brand new Llama Nemotron fashions.

The Nemotron and Cosmos Nemotron fashions will likely be obtainable in three parameter sizes — Nano, Super, and Ultra. Nano is essentially the most cost-effective mannequin constructed with low latency as the first focus. Super is a high-accuracy mannequin that may be run on a single GPU. Finally, Ultra is the highest-accuracy mannequin designed for information centre-scale purposes.

Nvidia highlighted that enterprises can entry the Nemotron mannequin household as downloadable fashions and as NIM. These fashions will even be obtainable as utility programming interfaces (APIs). While the fashions are open-source, they’re solely obtainable for tutorial and analysis utilization.