What this release signals
In 2026, Kubernetes continues to specialize. The 1.36 release reinforces the shift toward a cluster that better understands the types of workloads it runs, particularly in cloud-native and AI scenarios.
- DRA (Dynamic Resource Allocation) matures and positions itself as the foundation for GPU-intensive and specialized hardware workloads.
- Security advances with stronger isolation and more explicit policy controls, especially in multi-tenant environments.
- Operations and observability become increasingly native, reducing reliance on external tooling.
Headline changes in v1.36
Released on April 22, 2026, Haru delivers 70 enhancements: 18 stable, 25 in beta, and 25 in alpha.
- Workload Aware Scheduling (WAS) improves scheduling decisions for distributed workloads.
- DRA continues evolving for AI/HPC use cases, with a focus on machine learning clusters.
- Removals and deprecations (such as
gitRepoandexternalIPs) reduce potential attack surfaces.
For a deeper technical walkthrough, read the full analysis on this portal.
The name Haru reflects a Kubernetes that is becoming more predictable and better suited for modern workloads.