Red Hat, Inc., a global leader in open source solutions, has announced the general availability of Red Hat Enterprise Linux (RHEL) AI in hybrid cloud. This new foundational model platform allows users to efficiently develop, test, and run generative artificial intelligence (AI) models, improving enterprise applications. RHEL AI combines Granite, a family of open source language models, with InstructLab model alignment tools, based on the LAB methodology (Large Scale Alignment for Chatbots). The platform is packaged as an optimized RHEL image, ready to deploy on individual servers in hybrid cloud environments.
Despite the great potential of generative AI, the costs associated with acquiring, training, and fine-tuning large language models (LLMs) can be extremely high, with some models costing almost $200 million just for training. These costs do not include customization to specific organizational requirements, which often requires specialized data scientists and developers. Alignment to tailor models to specific data and business processes is crucial, making efficiency and agility essential for implementing AI in production.
Red Hat anticipates that in the next decade, smaller, efficient, and personalized AI models will become a significant part of the enterprise IT stack, alongside native cloud applications. To achieve this goal, generative AI must be accessible in terms of cost, availability, and compatibility with hybrid cloud. Red Hat maintains that an open source approach, similar to what has solved complex software problems for decades, can reduce barriers to effective adoption of generative AI.
An Open Source Approach to Generative AI
These are the challenges that RHEL AI aims to address, making generative AI more accessible, efficient, and flexible for CIOs and enterprise IT organizations across the hybrid cloud. RHEL AI helps:
- Empower generative AI innovation with Granite enterprise models, licensed as open source and aligned with a wide variety of generative AI use cases.
- Optimize the adaptation of generative AI models to business requirements with the InstructLab tool, which allows subject matter experts and developers in an organization to contribute unique skills and knowledge to their models, even without extensive data science knowledge.
- Train and deploy generative AI anywhere in the hybrid cloud by providing all the necessary tools to fine-tune and deploy models for production servers wherever the associated data lives. RHEL AI also provides a path to access Red Hat OpenShift AI to train, fine-tune, and serve these models at scale using the same tools and concepts.
RHEL AI is also backed by the benefits of a Red Hat subscription, which includes trusted enterprise product distribution, 24×7 production support, extended model lifecycle support, and legal protections of Open Source Assurance.
RHEL AI extending across the hybrid cloud
Bringing a more consistent foundational model platform to where an organization’s data resides is crucial to support production AI strategies. As an extension of Red Hat’s hybrid cloud platforms, RHEL AI will span nearly every imaginable enterprise environment, from on-premise data centers to the network edge and public cloud. This means that RHEL AI will be available directly from Red Hat, from Red Hat’s original equipment manufacturer (OEM) partners, and to run on the world’s largest cloud providers, including Amazon Web Services (AWS), Google Cloud, IBM Cloud, and Microsoft Azure. This allows developers and IT organizations to harness the power of hyperscale computing resources to build innovative AI concepts with RHEL AI.
Availability
RHEL AI is generally available through the Red Hat Customer Portal to run on-premise or to deploy on AWS and IBM Cloud as a “bring your own subscription” (BYOS) offering. Availability of a BYOS offering on Azure and Google Cloud is planned for the fourth quarter of 2024, and RHEL AI is also expected to be available on IBM Cloud as a service later this year.
Red Hat plans to expand the availability of RHEL AI through cloud partners and OEMs in the coming months, providing even more options in hybrid cloud environments.