NVIDIA Introduces Llama Nemotron: A New Generation Of Open Reasoning AI Models To Power Agentic AI Platforms

NVIDIA has launched the Llama Nemotron family of open reasoning AI models, designed to build advanced agentic AI platforms. These models, post-trained by NVIDIA, offer up to 20% higher accuracy and 5x faster inference speed than leading alternatives. Available via NIM microservices, the models come with developer tools such as AI-Q Blueprint and AgentIQ to support custom AI agent development.

NVIDIA has launched a new family of open-source reasoning AI models — Llama Nemotron — designed to serve as a robust foundation for building agentic AI platforms across industries. Announced at GTC, this strategic release marks a significant step in the evolution of enterprise AI, providing developers and organisations with models that can solve complex problems through advanced reasoning, multistep math, coding, and decision-making capabilities.

The Llama Nemotron family is built on Meta’s Llama models and post-trained by NVIDIA to improve performance, accuracy, and inference speed. These models are tailored to support businesses in deploying autonomous AI agents that operate independently or in collaboration, offering increased functionality, cost efficiency, and operational scalability.

Enhanced reasoning and performance through post-training

The post-training conducted by NVIDIA has refined the models’ accuracy by up to 20% over the base models and increased inference speed by five times compared to other leading open reasoning models. This results in more efficient handling of advanced reasoning tasks, from solving technical problems to making enterprise decisions.

NVIDIA performs this post-training using its DGX™ Cloud infrastructure, applying synthetic data generated by NVIDIA Nemotron™ and other open models, along with curated datasets cocreated by NVIDIA. These training resources, including datasets and optimisation techniques, are made openly available to developers, encouraging customisation and transparency in AI development.

The Llama Nemotron models are available as NVIDIA NIM™ microservices, offered in Nano, Super, and Ultra sizes. The Nano model is best suited for PCs and edge devices, Super for single GPU systems, and Ultra for multi-GPU servers that require maximum reasoning performance.

Industry collaboration on Agentic AI platforms

A number of global companies are collaborating with NVIDIA to integrate the new models into their platforms and services. These include:

• Microsoft is incorporating Llama Nemotron and NIM microservices into its Azure AI Foundry, enhancing services like Azure AI Agent Service for Microsoft 365.
• SAP is using the models to improve Joule, its AI copilot, and SAP Business AI solutions. According to Walter Sun, Global Head of AI at SAP, “We are collaborating with NVIDIA to integrate Llama Nemotron reasoning models into Joule to enhance our AI agents, making them more intuitive, accurate and cost effective.”
• ServiceNow is adopting the models to enhance enterprise productivity across sectors with higher-performing AI agents.
• Accenture is leveraging the models within its AI Refinery platform, enabling clients to create custom AI agents for industry-specific challenges.
• Deloitte plans to implement Llama Nemotron in its Zora AI platform, aiming to support agents that can replicate human decision-making with deep industry knowledge and transparency.

Other collaborators include Amdocs, Atlassian, Box, Cadence, CrowdStrike, and IQVIA, all of whom are developing advanced agentic AI systems powered by the new models.

Developer access and AI tools for enterprise integration

The new models are accessible via build.nvidia.com and Hugging Face. Developers in the NVIDIA Developer Program can use them for free for development, testing, and research. For production use, the models are available with NVIDIA AI Enterprise, optimised for deployment in cloud and data centre infrastructure.

To support developers and enterprises building agentic AI platforms, NVIDIA offers several new tools:

• NVIDIA AI-Q Blueprint: An architecture that enables AI agents to autonomously perceive, reason and act. It incorporates NeMo Retriever™ for multimodal information access and the AgentIQ toolkit for agent-data connection, optimisation and transparency. The Blueprint is expected to be released in April 2025.
• NVIDIA AI Data Platform: A reference design for enterprise infrastructure supporting AI query agents built on the AI-Q Blueprint.
• NVIDIA NeMo microservices: These enterprise-grade tools establish a continuous data learning cycle, allowing AI agents to improve via feedback. Developers can build a data flywheel — a dynamic system of feedback and improvement — with this platform, optimising enterprise-level reasoning capabilities.

Accelerating industry with reasoning AI

The adoption of agentic AI across industries is gaining momentum, and NVIDIA’s Llama Nemotron family provides the foundational models necessary to drive this shift. By offering performance-optimised, open-source reasoning models along with essential developer tools and microservices, NVIDIA is enabling a new generation of intelligent agents capable of complex thought, collaboration, and action.

As Jensen Huang, founder and CEO of NVIDIA, states, “Reasoning and agentic AI adoption is incredible. NVIDIA’s open reasoning models, software and tools give developers and enterprises everywhere the building blocks to create an accelerated agentic AI workforce.”

About NVIDIA

Founded in 1993, NVIDIA is a global leader in accelerated computing. Initially revolutionising the gaming industry with its powerful graphics processing units (GPUs), the company has since transformed industries including healthcare, automotive, and finance through innovations in AI, deep learning, and high-performance computing. NVIDIA’s AI platforms are widely recognised for powering complex applications, from robotics and medical imaging to scientific research and autonomous vehicles.

Today, NVIDIA continues to push the frontiers of computing, focusing on technologies that empower developers, scientists, and enterprises to solve the world’s most pressing challenges. With its robust ecosystem of hardware, software, and AI frameworks, NVIDIA remains a central figure in shaping the future of artificial intelligence and intelligent computing worldwide.