Red Hat is introducing a unified AI platform for deploying and managing AI models, agents, and applications.
Red Hat AI Enterprise aims to help companies that are stuck in AI pilot phases as a result of fragmented tools and infrastructure, by offering a standard environment where organizations can perform AI inference, customize and tune models, and deploy and manage agents, using any model or hardware.
It leverages optimized runtimes for inference, such as vLLM and the llm-d framework, to provide high-throughput and low-latency model serving, as well as providing a standardized API layer via the Llama Stack and supporting the Model Context Protocol (MCP) to allow connection to external systems.
Additionally, to enable greater trust in AI, Red Hat AI Enterprise offers tools for drift detection, bias monitoring, and model explainability, including evaluation frameworks like RAGAS, which measure the quality of RAG-based systems.
According to Red Hat, while the platform is important in laying the foundations for AI implementation, its true strength is in its Day-2 capabilities, such as dynamic resource scaling, integrated monitoring, unified security, and rolling updates to prevent downtime.
“Red Hat AI Enterprise is more than just a collection of tools—it is a strategic foundation for the AI-driven era. By bridging the gap between experimentation and production, it helps organizations innovate faster while maintaining the security posture and control required by the modern enterprise,” Jennifer Vargas, principal product marketing manager at Red Hat, wrote in a blog post.
To coincide with the launch of Red Hat AI Enterprise, the company also announced a new version of Red Hat AI, its platform for building and running AI applications.
Red Hat AI 3.3 features an expanded model ecosystem that includes production-ready, compressed versions of Mistral-Large-3, Nemotron-Nano, and Apertus-8B-Instruct. It also offers support for additional hardware, such as support for AMD MI325X accelerators, an expansion to the hardware certification for NVIDIA’s Blackwell Ultra, and a technology preview for supporting Intel CPUs.
In addition, this release includes a technology preview of Models-as-a-Service (MaaS), enabling IT teams to set up self-service access to privately hosted models, promoting increased AI adoption within the enterprise.
Other key updates in Red Hat AI 3.3 include:
- The Red Hat AI Python Index, which provides access to enterprise-grade versions of critical tools like Docling, SDG Hub, and Training Hub
- Real-time telemetry across AI workloads, llm-d deployments, and MaaS cluster and model usage
- A technology preview of a NeMo Guardrails integration that allows developers to better enforce operational safety and alignment
