It will be an understatement to say, that AI has dominated the overall tech field for some time now. Global giants are racing to launch AI enabled hardware, software at breakneck pace, and CES 2025 is no exception. Perhaps among the most awaited keynotes at this year’s CES, Nvidia — apart from the Blackwell-powered RTX GPUs — has also unveiled its latest Nemotron Model Families. These models, including the Llama Nemotron and Cosmos Nemotron families, aim to aid in the growth of agentic AI (wherein specialized AI agents collaborate to tackle complex tasks and automate repetitive processes).
The Llama Nemotron family of large language models (LLMs) is at the core of Nvidia’s agentic AI initiative. Built on the Llama foundation models, these open-source solutions are optimized for computational efficiency, and can perform a wide range of tasks, including coding, mathematical computations, and function calling. Llama Nemotron models leverage NVIDIA’s latest pruning techniques and high-quality datasets to cater to the need for scalable AI solutions, enabling developers to design and deploy intelligent agents on diverse NVIDIA-accelerated platforms. So far, enterprises like SAP and ServiceNow have already expressed interest in integrating Llama Nemotron models into their systems.
“Artificial intelligence is entering a new era — agentic AI — where teams of specialized agents can help people solve complex problems and automate repetitive tasks,” Nvidia announced in a blog post. “With custom AI agents, enterprises across industries can manufacture intelligence and achieve unprecedented productivity. These advanced AI agents require a system of multiple generative AI models optimized for agentic AI functions and capabilities. This complexity means that the need for powerful, efficient, enterprise-grade models has never been greater.”
“Agentic AI is the next frontier of AI development, and delivering on this opportunity requires full-stack optimization across a system of LLMs to deliver efficient, accurate AI agents,” said Ahmad Al-Dahle, VP and head of GenAI at Meta. “Through our collaboration with NVIDIA and our shared commitment to open models, the NVIDIA Llama Nemotron family built on Llama can help enterprises quickly create their own custom AI agents.”
Nvidia enters the realm of ‘World Models’
In addition to this, Nvidia has also introduced the Cosmos Nemotron vision language models (VLMs). These models are designed to analyze and respond to visual inputs, making them indispensable for applications in fields such as robotics, retail, healthcare, and sports analytics. They enable developers to create agents capable of processing images and videos from autonomous machines, warehouses, and live events.
This comes alongside the introduction of the Cosmos World Foundation Models (WFMs), which focus on generating physics-aware simulations and videos. These models are categorized into three tiers—Nano, Super, and Ultra.
Nvidia claims that Cosmos WFMs have been trained on a dataset encompassing 20 million hours of video and environmental interactions, and can generate synthetic data for robotics and autonomous vehicle applications. Developers can use these models to simulate real-world scenarios, as well as “generate photoreal videos from controlled 3D scenarios.” Both Nemotron and Cosmos models have been made openly available through platforms like GitHub, Hugging Face, and its NGC catalog.
