NVIDIA unveils the Nemotron series of large language models, aiming to accelerate the emergence of agent-based AI.

21st Century Business Report's White Yang in Beijing

At the CES 2025 conference today, Nvidia founder and CEO Jensen Huang unveiled the new Llama Nemotron series of large language models.

Huang said that artificial intelligence is entering a new era - agent AI, where professional AI agents can help people solve complex problems and automatically execute repetitive tasks.

He further stated that with custom AI agents, businesses in all industries can achieve unprecedented productivity. However, these advanced AI agents require a system of multiple generative AI models optimized for agent AI functions and capabilities. This complexity means the demand for powerful and efficient enterprise-level models has never been stronger.

Nvidia's Llama Nemotron models, which have Nano, Super, and Ultra versions, are designed to help developers create and deploy AI agents in a variety of applications, including customer support, fraud detection, and product supply chain and inventory management optimization.

The models are based on Llama and have been optimized for performance, with Nvidia's latest technologies and high-quality datasets. They excel in instruction tracking, chat, function calls, coding, and math, while being optimized for size to run on Nvidia's extensive acceleration computing resources.

The Llama Nemotron models will be available as downloadable models and NVIDIA NIM microservices, making it easy to deploy in the cloud, data centers, PCs, and workstations.

Businesses can also use NVIDIA NeMo microservices to customize models for their specific use cases and domains, simplifying data management, accelerating model customization and evaluation. Additionally, developers can integrate retrieval-enhanced generation capabilities using NVIDIA NeMo Retriever, connecting models to their enterprise data.