Red Hat AI 3.5 targets the operational gap between successful AI pilots and accountable, production-scale deployments
Red Hat has announced the general availability of Red Hat AI 3.5, adding safety, observability, governance and workload-management capabilities aimed at helping enterprises move AI beyond experimentation and into controlled production environments.
The release reflects a broader shift in enterprise AI priorities. As organizations deploy more models and agentic applications, the challenge is no longer simply accessing AI capabilities, but establishing the controls needed to manage risk, infrastructure consumption, performance and accountability at scale.
Red Hat AI 3.5 introduces EvalHub for pre-deployment model evaluation, including safety benchmarking and compliance reporting. New observability dashboards provide visibility into inference health, GPU utilization and model performance, while per-user token metering supports usage transparency and showback.
“The conversation has moved from getting AI into production to running it at scale as trusted enterprise infrastructure, which requires safety evidence, governed agents, cost attribution and multi-tenancy,” said Joe Fernandes, vice president and general manager, AI Business Unit, Red Hat.
Multi-tenancy is another significant focus. Fair-share GPU scheduling and priority-aware serving are designed to balance competing workloads on shared infrastructure, while hosted control planes and OpenShift Virtualization provide stronger isolation for environments serving multiple tenants.
The platform also strengthens the enterprise agentic AI stack. AutoRAG connects enterprise data with agentic applications, while agent templates and starter kits provide preconfigured patterns for tasks such as code review, document processing and research workflows. Inference-Time Scaling can dynamically adjust compute according to query complexity, potentially improving GPU economics.
For CIOs and platform leaders, Red Hat AI 3.5 signals a move toward treating AI as governed enterprise infrastructure rather than a collection of isolated projects. The emphasis on safety evidence, observability, resource controls and agent governance could become increasingly important as AI workloads become deeply embedded in business operations.