NVIDIA announced the Llama Nemotron Family (LLMS) on Monday. With the rise of artificial intelligence (AI) agents, new, more complex AI models are needed to handle the workflow of AgentIC AI, the company said. The tech giant highlighted the need for more power and higher efficiency, saying the Nemotron family model can create and deploy AI agents in a variety of applications. The company claims that the AI model will be available to enterprises through NVIDIA NIM microservices.
NVIDIA introduces Nemotron family AI model
The tech giant announced a new series of open source LLMSs in a blog post called Nemotron. The series also includes the Cosmos Nemotron Vision Language Model (VLM), which can be used to build AI agents that analyze and respond to images and videos. NVIDIA said vision-focused agents can be deployed in automated machines, hospitals, stores and warehouses, as well as sports events, movies and news.
The NVIDIA LLAMA NEMOTRON model is built by Meta’s Llama Foundation model and is said to be optimized to build and develop AI agents. Although the company did not disclose architectural and technical details, it claims that the models were trained using “the latest technology and high-quality datasets.” These models can be used to train proxy functions such as the following instructions, chat, function calls, encoding and math, and more. Nemotron is said to also optimize the size of the AI agent to make it easy to deploy.
NVIDIA said SAP, ServiceNow and other AI proxy platform providers will be one of the first to use the new Llama Nemotron model.
The Nemotron and Cosmos Nemotron models will be divided into three parameter sizes – Nano, Super and Ultra. Nano is the most cost-effective model with a main focus on low latency. Super is a highly accurate model that can run on a single GPU. Finally, Ultra is the most accurate model designed for data center-scale applications.
NVIDIA emphasizes that businesses can access the Nemotron model family as downloadable models and NIMs. These models will also be provided as application programming interfaces (APIs). Although the models are open source, they are only used for academic and research usage.
