Red Hat Launches Unified ‘Metal-to-Agent’ AI Platform and Unveils AI 3.3, Deepening NVIDIA Collaboration

Overview

• Red Hat introduced Red Hat AI Enterprise, a stack that links Red Hat Enterprise Linux and OpenShift with high‑performance inference, model tuning, and agent deployment across environments.
• The 3.3 release broadens model options with validated compressed builds of Mistral‑Large‑3, Nemotron‑Nano, and Apertus‑8B‑Instruct, and enables newer choices like Ministral 3 and DeepSeek‑V3.2 with multimodal gains such as faster Whisper.
• Red Hat and NVIDIA co‑engineered the Red Hat AI Factory, combining Red Hat AI Enterprise with NVIDIA AI Enterprise to speed production on NVIDIA infrastructure with preconfigured NIM microservices and NeMo agent lifecycle tools.
• Hardware support expands with a technology preview for generative AI on Intel CPUs, added certification for NVIDIA Blackwell Ultra, and support for AMD MI325X accelerators.
• Operational features include a technology preview of Models‑as‑a‑Service, integrated telemetry for AI workloads with a preview of NeMo Guardrails for safety, and GPU pooling with checkpointing to protect long‑running jobs.