At the recent GTC 2025 conference, NVIDIA introduced the Llama Nemotron family of AI models, designed to empower developers and enterprises in building advanced AI agents capable of independent reasoning and complex decision-making. The Llama Nemotron models are built upon Meta's Llama architecture and have undergone extensive post-training by NVIDIA to enhance their capabilities in multistep mathematics, coding, and intricate reasoning tasks. This refinement has resulted in a 20% increase in accuracy and a fivefold improvement in inference speed compared to previous models.

Tailored for Diverse Deployment Needs

Recognising the varied requirements of enterprises, NVIDIA offers the Llama Nemotron models in three distinct sizes:

Nano: Optimised for PCs and edge devices, providing high accuracy in compact deployments.

Super: Designed for single GPU instances, balancing throughput and precision.

Ultra: Engineered for multi-GPU servers, delivering maximum agentic accuracy for complex tasks.

These models are accessible through NVIDIA's NIM microservices platform, facilitating seamless integration into existing infrastructures.

Collaborations with Industry Leaders

NVIDIA's initiative has garnered support from prominent organisations such as Microsoft, SAP, ServiceNow, Accenture, and Deloitte. These collaborations aim to leverage the Llama Nemotron models to enhance their AI offerings, streamline operations, and drive innovation across various sectors.

Click to read AI news here!

Empowering Developers with Open Resources

In a move to foster transparency and collaboration, NVIDIA is releasing the tools, datasets, and post-training optimisation techniques used in developing the Llama Nemotron models. This open approach enables enterprises to customise and build their own reasoning models, tailored to specific business needs.

Read: Goodbye Chaos? Mumbai’s Public Transport Gets a Makeover

A Vision for the Future of Work

Jensen Huang, CEO of NVIDIA, emphasised the transformative potential of these models: "NVIDIA’s open reasoning models, software, and tools provide the building blocks for developers and enterprises to create an accelerated agentic AI workforce." With the introduction of the Llama Nemotron family, NVIDIA is poised to redefine the capabilities of AI agents, bringing human-like reasoning to machines and paving the way for a new era in enterprise automation.