New features, improvements, and changes for Container Service for Kubernetes (ACK) in 2026.
Background
-
See Version Guide for Kubernetes versions supported by Container Service for Kubernetes (ACK).
-
Container Service for Kubernetes (ACK) supports the following operating systems: ContainerOS, Alibaba Cloud Linux 4 Container-optimized, Alibaba Cloud Linux 3 Container-optimized, Alibaba Cloud Linux 3, Alibaba Cloud Linux 3 for Arm, Alibaba Cloud Linux UEFI 3, Red Hat, Ubuntu, and Windows. See Operating systems.
March 2026
|
Product |
Feature |
Description |
Related documentation |
|
Container Service for Kubernetes |
Support for Alibaba Cloud Linux 4 Container-optimized images |
Alibaba Cloud Linux 4 Container-optimized is a container-optimized OS based on Alibaba Cloud Linux 4. It ships with the ANCK 6.6 kernel and enables cgroup v2 by default for finer-grained resource management. |
|
|
Deploy agent applications as Knative services with AgentScope |
Deploy AgentScope agent applications to ACK Knative to use serverless capabilities such as auto scaling (including scale-to-zero) and revision management. |
Use AgentScope to deploy agent applications as Knative services |
|
|
Deploy a Qwen large language model inference service in an Auto Mode cluster |
Deploy the Qwen3.5-4B large language model as a serverless inference service with Auto Mode node pools and Knative Serving, eliminating manual GPU management. |
||
|
ACK Argo Workflow Cluster |
Build large-scale Argo Workflows with the Python SDK |
The Hera Python SDK lets you define Argo Workflows in Python instead of YAML, with support for complex tasks, easy testing, and Python ecosystem integration. |
|
|
Event-driven CI pipelines with EventBridge |
Build event-driven CI pipelines with EventBridge and Argo Workflows to simplify and accelerate application delivery. |
||
|
Build a CI pipeline for a Go project with a workflow cluster |
A workflow cluster is a fully managed, auto-scaling service built on Argo Workflows with zero O&M overhead for CI pipelines. |
Build a CI pipeline for a Go project with a workflow cluster |
|
|
ACK Edge |
Expose services by using Gateway with Inference Extension |
Use Gateway with Inference Extension in an ACK Edge cluster to configure HTTP routing, modify request headers, and split traffic by weight. |
February 2026
|
Product |
Feature |
Description |
Related documentation |
|
Container Service for Kubernetes |
ACK AI Assistant introduces an agent mode |
ACK AI Assistant adds an agent mode for root cause analysis of complex faults, best practice recommendations, and simplified cluster O&M. |
|
|
ContainerOS v3.7 is now available |
ContainerOS v3.7 adds an image with pre-installed GPU drivers. |
||
|
Kubernetes 1.35 is now available |
ACK now supports Kubernetes 1.35, which drops cgroup v1 support. The node OS must support cgroup v2 (stable since Kubernetes 1.25), or the kubelet fails to start. |
||
|
Deploy and route LLM services in Knative with an AI inference gateway |
Gateway with Inference Extension, based on the Kubernetes Gateway API, pairs with Knative to simplify generative AI inference management. It provides L7 routing, load balancing across inference workloads, and elastic GPU scaling based on request concurrency. |
Deploy and intelligently route LLM services by using an AI inference gateway |
|
|
Add custom admission checks for ACK Kube Queue |
By default, ACK Kube Queue checks only the |
||
|
ACK Argo Workflow Cluster |
ACK Argo Workflow Cluster is now available |
ACK Argo Workflow Cluster is a fully managed, zero-O&M service for efficient and stable large-scale workflow orchestration. |
January 2026
|
Product |
Feature |
Description |
Related documentation |
|
Container Service for Kubernetes |
Support for dynamic volumes with CPFS for AI Computing |
Dynamic volumes automate on-demand storage provisioning for CPFS for AI Computing, eliminating manual PV management. Multiple applications can concurrently read and write to share code, configuration files, and intermediate data. |
|
|
Support for on-demand data prefetching from OSS to high-performance volumes |
Prefetch cold data from Object Storage Service (OSS) to a high-performance volume, such as CPFS for AI Computing or a cloud disk, before AI training or data analytics tasks. The volume is automatically reclaimed after task completion. |
Prefetch data from Object Storage Service (OSS) to a high-performance volume on demand |
|
|
Support for scheduling applications to a hybrid cloud node pool |
A hybrid cloud node pool registers data center (IDC) nodes with an ACK cluster for unified management and scheduling of on-premises and cloud resources. |
||
|
yurt-hub component is now available |
yurt-hub is a node component that provides autonomy for hybrid cloud node pools in ACK clusters. |