Red Hat Releases Red Hat AI 3.5 with Enhanced Multi-Tenancy and Observability

Red Hat has released Red Hat AI 3.5, an update to its enterprise AI platform designed to bring the same operational rigor used for business-critical infrastructure to AI workloads. The release, announced on October 14, 2026, strengthens multi-tenancy support for AI service providers, including full software and hardware isolation for workloads handling sensitive data, proprietary models, and regulated information.

The new version introduces priority-aware serving on shared GPU infrastructure, dynamically allocating GPU capacity based on workload priority so critical business tasks can preempt lower-priority jobs while background workloads use idle capacity. It also adds EvalHub for pre-deployment model validation, supporting risk-focused safety benchmarking and regulatory compliance, along with a real-time observability dashboard for inference health, GPU utilization, and AI model performance. Non-admin users can view per-user token consumption metering and distributed inference workloads.

Shared GPU controls for multi-tenant inference provide cross-tenant resource allocation, admission control, and priority-based request routing. CPU offloading is now generally available and storage offloading is in developer preview, enabling models to handle longer conversations and larger documents without additional GPU hardware. Red Hat AI Hub also offers agent templates and starter kits for common enterprise scenarios such as code review, document processing, and research workflows.

Tushar Katarki, senior director of AI products at Red Hat, said the release provides the necessary operational guardrails, verifiable trust, and multi-tenancy controls for AI to run as a business-critical service rather than an unpredictable experiment.