OpenShift sandboxed containers 1.13 moves GPU-backed confidential AI to GA while a new Agent Sandbox preview isolates code that autonomous agents generate.

A new IBM lab blueprint joins DCGM workload signals, cloud cost data and alerts-as-code so platform teams can separate useful accelerator work from paid-for idle capacity.

Red Hat and vLLM engineers combine token-parallel prefill, sharded KV cache and host-memory decode to fit contexts that standard parallelism cannot hold.

MiniMax-M3.1-Flash arrives with sparse-attention, MoE and speculative-decoding paths for NVIDIA and AMD accelerators, plus reproducible recipes.

Red Hat’s OpenShift walkthrough separates browser login from server-side token introspection and documents the operator-era replacement for legacy SSO.

An OpenShift pattern combines gateway isolation, network policy, identity-aware tool permissions and rate limits instead of treating authentication as sufficient.

Red Hat’s walkthrough shows how to trade Keycloak’s SSO layer for direct directory binding—and where TLS becomes non-negotiable.

RHEL CoreOS 10 is a Technology Preview in OpenShift 4.22; Red Hat’s planned mixed-stream model would let administrators validate and move nodes gradually.

The clinical-simulation project joins OpenShift orchestration, IBM Fusion storage and NVIDIA inference in a permission-aware, on-premises architecture.

The release gives Kubernetes operators explicit control over pod metadata, adds standards-based OCSF export and hardens failure paths around sandbox policy enforcement.

A new engineering guide joins Podman, Red Hat AI Inference, GuideLLM and Open WebUI into a repeatable local evaluation loop.

A four-stage method moves OpenShift chaos testing from one-off pod failures to measurable workflows that teams can replay before releases and infrastructure changes.

The reference deployment lets an agent inspect current Azure resource state through MCP while OPA, read-only credentials and audit logs constrain its reach.

Red Hat and NEC built a shared inventory and validation framework for more than 200 certificate-related artifacts across the platform.

The open-source plug-in captures and edits model internals during CUDA Graph replay, avoiding the eager-mode penalty that keeps many safety tools out of production.

Red Hat’s new benchmark replaces Python-heavy toy problems with reproducible Go, Python and C tasks drawn from Podman, etcd, Ansible, KubeVirt and related projects.

The first beta moves the Java baseline, networking stack, serialization layer and persistence defaults at once, making early testing more than a compatibility exercise.

A production case study shows how queue-aware and prefix-cache-aware routing can turn idle accelerator capacity into predictable training-data output.

The new open-source library separates quantization descriptions from GPU execution choices, then compiles a tuned kernel for each workload.

A first-party reference design joins simulation, evaluation, signed OCI artifacts and fleet delivery on one AI workstation.

A troubleshooting-assistant design replaces flat fan-out with layered orchestrators, trading additional model calls for narrower routing decisions and contained failures.

Red Hat’s two-site OpenShift lab preserved OIDC sessions during site failure, but the Keycloak 26.7 stateless work remains a preview that operators must validate.

Red Hat says moving back to 26.4 can fail after Kubernetes records v2beta1 as a stored version, so operators need a tested rollback path before upgrading.

The validated architecture joins IBM’s sovereignty control plane and integrated infrastructure on Red Hat OpenShift, targeting regulated enterprises and service providers.

Clusters moving from a 4.21 multi-architecture payload should verify the payload type and avoid the affected 4.22.15–4.22.17 targets; the Machine Config Operator fix is present from 4.22.18 onward.

The partners say they will validate new MiniMax serving paths before weights arrive, then feed production lessons back into upstream vLLM.

IBM’s third-quarter disclosure points to continued momentum in its hybrid-cloud business, but the segment reporting stops well short of showing which Red Hat products drove it.

The validation gives platform teams a common Kubernetes control plane for GPU-backed virtual machines and containerized AI workloads, but Red Hat still advises workload-specific testing.

The GitOps-managed blueprint coordinates identity, secrets, runtime controls, supply-chain evidence and compliance without pretending one product can deliver zero trust.

A new FlashSystem blueprint gives Ramen and Advanced Cluster Management control of workload recovery while storage teams retain the replication layer.

Day-zero support for Recursive’s sparse model shows how model-specific kernels can turn long-context serving into a single-Blackwell deployment.

The automaker says Ansible cut one large configuration workflow from 60 hours to 20 minutes while replacing a 30-year-old management system.

The messaging specialist can now participate in multi-agent workflows while running against managed Red Hat AI Inference or self-managed OpenShift AI, RHEL AI and vLLM.

The Singapore centre will pair reusable architectures and upstream contributions with a government co-engineering project spanning AI, cloud-native systems and day-two operations.

A new engineering blueprint combines pull-based reconciliation, model-aware telemetry, staged delivery and local rollback for AI fleets that cannot depend on cloud-style connectivity.

Red Hat’s engineering guide uses typed probabilities for messy text while keeping thresholds, authorizations and execution in ordinary application policy.

The OpenShift AI reference build combines routed specialist agents, MCP actions, persistent sessions and human handoffs with DeepEval, OpenTelemetry and MLflow.

Red Hat’s migration guide says legacy Scalprum mounts can stop the portal at startup, while standalone OCI plug-in images and Backstage extension blueprints redraw the packaging boundary.

A reference project puts Artifactory, Podman and Red Hat dependency analytics on one workstation so teams can test remediation before involving shared infrastructure.

The expanded Red Hat AI Factory architecture combines OpenShell runtime isolation, governed model access and Day 0 NVIDIA hardware enablement, but native OpenShell integration is still being built.

Red Hat limits its hardware-enablement delta to the kernel, QEMU and libvirt, with a planned transition back to mainline RHEL once the patches mature upstream.

The launch reaches beyond one serving engine, tying vLLM and llm-d inference to Hugging Face, TorchTitan, VeRL and Ray on the same open TPU foundation.

Linux Foundation research sponsored in part by Red Hat pairs reported developer time savings with rising review load, weak contribution policies and burnout risk.

The airline’s 1,200-developer cloud workspace platform moves agent compute off laptops while keeping tools, secrets and nested containers inside governed project environments.

A new llm-d flow-control path prioritizes workloads only when a model-serving pool saturates, with per-tenant fairness and explicit limits for long outputs.

A merged chunk/noisy-KV path brings stage-pipelined autoregressive video diffusion into the upstream serving stack, but it remains explicitly experimental.

The release adds FP8 hidden-state storage, a faster fused loss path, Mooncake transfer changes and live evaluation tooling for teams building speculative decoders.

The first major release aligns core components, graduates the router API to v1 and adds flow control, high availability and coordinated rollouts for disaggregated serving.

The deployment guide fits Reflection Beam on four Blackwell or eight Hopper GPUs, but relies on an unreleased vLLM build and CUDA-only kernels.

NIXL EP and reusable CUDA graphs cut measured topology-change pauses from tens of seconds to less than two, but orchestration integration is still in progress.