**NVIDIA Unveils Llama Nemotron Models for Enhanced AI Reasoning** During GTC, NVIDIA introduced its open Llama Nemotron family of models, engineered to equip developers and businesses with advanced AI reasoning capabilities. These models allow for the creation of AI agents that can function independently or collaboratively to tackle intricate tasks. Based on Llama models, the Llama Nemotron family significantly boosts AI reasoning through enhanced post-training processes aimed at improving multistep math, coding, and decision-making accuracy by up to 20% and optimizing inference speed by 5x compared to other leading models. As a result, enterprises can expect better decision-making capabilities and reduced operational costs. Major industry collaborators like Accenture, Microsoft, SAP, and Deloitte are integrating these models into their platforms to enhance AI functionalities. For instance, Microsoft is embedding Llama Nemotron models in its Azure AI Foundry, while SAP is utilizing them to improve AI solutions for its Joule copilot. The Llama Nemotron models come in three sizes—Nano, Super, and Ultra—each tailored for specific deployment requirements.

These models are accessible as NVIDIA NIM™ microservices, with the Nano model optimal for PCs and edge devices, the Super model excelling on single GPUs, and the Ultra model designed for multi-GPU servers. NVIDIA plans to share the tools and datasets used for model training openly, allowing enterprises to create tailored reasoning models. The deployment of these models is streamlined through NVIDIA’s AI Enterprise suite, featuring essential tools like the AI-Q Blueprint and customizable data platforms. The Llama Nemotron Nano and Super models are currently available as APIs, while enterprises can run them in production with NVIDIA AI Enterprise. The forthcoming AI-Q Blueprint is expected in April, and developers can access the NVIDIA AgentIQ toolkit now on GitHub. NVIDIA continues to be a leader in accelerated computing, emphasizing its commitment to innovation, even as market conditions and technological landscapes evolve.