Bringing Nemotron models to the Red Hat AI Factory with NVIDIA

16 marzo 2026Rob Greenberg, Tyler Michael Smith2 minuti (tempo di lettura)

Following the successful launch of the Red Hat AI Factory with NVIDIA, Red Hat is pleased to announce the latest update in our collaboration with NVIDIA – delivering Day 0 support for the NVIDIA Nemotron open model family on the Red Hat AI Factory with NVIDIA. With this effort, we are providing a fully optimized, open source pathway for enterprise-grade generative AI.

From infrastructure to intelligence: Accelerating AI mainstream enterprise adoption

The Red Hat AI Factory with NVIDIA was designed to provide a turnkey environment for developing and deploying AI at scale. Today’s announcement expands this beyond the software and hardware stack, integrating NVIDIA’s high-performance foundation models directly into the hybrid cloud workflow.

As the industry sees the growth of proprietary, closed-box AI, Red Hat is doubling down on open source–from models to software and the surrounding ecosystems–to provide more choice and flexibility for enterprises. Our collaboration with NVIDIA helps ensure that the intelligence running within the Red Hat AI Factory with NVIDIA is transparent, reproducible, and fully optimized for the unique security and sovereignty needs of global enterprise and government customers.

Bringing NVIDIA Nemotron to the enterprise

NVIDIA Nemotron is a family of open models, data, and libraries designed to power transparent, efficient, and specialized agentic AI development across industries. Nemotron open models are optimized for reasoning, retrieval-augmented generation (RAG), and instruction-following tasks. They enable enterprises to accelerate AI innovation across sectors such as financial services, healthcare, and public sector implementations, supporting secure and scalable deployment of generative AI solutions.

To make sure these models are ready for mission-critical environments, we are continuing to deepen engineering workflows with NVIDIA to deliver Day 0 support for new NVIDIA Nemotron models, including Nemotron 3 Super, on vLLM and Red Hat AI. This means that customers will be able to immediately run NVIDIA Nemotron models on Red Hat AI at the moment of release.

As part of this effort, Red Hat will provide rigorous inference performance and accuracy benchmarks on certified GPUs to validate Nemotron models on Red Hat AI platforms, including Red Hat AI Enterprise, and provide clear deployment guidance. From there, Red Hat AI Enterprise acts as the engine for model delivery and validation within the Red Hat AI Factory with NVIDIA. This empowers organizations with enhanced performance and enterprise-ready packaging, with models delivered as OCI artifacts and Modelcars that enable container vulnerability scanning and consistent lifecycle management across the hybrid cloud.

The future is open

The Red Hat AI Factory with NVIDIA is a complete engine for enterprise innovation. By integrating NVIDIA Nemotron open models into this framework, we are giving customers a reliable, high-performance toolkit that they can truly own. Together, we are proving that the future of enterprise AI is open, secure, and built on a foundation of collaboration.

Get started today with the Red Hat AI Factory with NVIDIA and find us at NVIDIA GTC to speak to one of our experts directly.

Sugli autori

Rob Greenberg

Principal Product Manager

My name is Rob Greenberg, Principal Product Manager for Red Hat AI, and I came over to Red Hat with the Neural Magic acquisition in January 2025. Prior to joining Red Hat, I spent 3 years at Neural Magic building and delivering tools that accelerate AI inference with optimized, open-source models. I've also had stints as a Digital Product Manager at Rocketbook and as a Technology Consultant at Accenture.

Read full bio

Tyler Michael Smith

Chief Architect - Inference Engineering and Member of Technical Staff

Tyler received a PhD in Computer Science from The University of Texas at Austin, studying high performance dense linear algebra - microkernels, parallelism, and theoretical lower bounds on data movement. Later joined Neural Magic, first working on a graph compiler for sparse neural network inference on CPUs. Now he works on large language model inference in vLLM and llm-d, especially in large scale serving.

Read full bio