About GKE modes of operation

This page helps you to choose the Google Kubernetes Engine (GKE)mode of operationthat's the best fit for your requirements. This information is intended for thefollowing people:

  • Application developers who want to run containerized workloads optimallybased on workload requirements.
  • Platform administrators who want to know about the recommended methods to useGKE.

If you want to learn about whether GKE as aplatform is thebest choice for your containerized applications, refer to theGKE overviewandGKE and Cloud Run.

Introduction to Autopilot mode and Standard mode

GKE lets you choose one of the followingmodes of operation,which determines factors like who manages the infrastructure, how you're billedfor resources, and how much flexibility you have:

  • Autopilot (recommended): a managed experience in whichGKE handles most of your infrastructure for you so that youcan focus on your applications. Autopilot is optimized for mostproduction scenarios and implements many of Google's best practices forsecurity, scaling, and workloads.

    For more information, seeAbout GKE Autopilot.

  • Standard: a flexible experience in which you have directcontrol over your node infrastructure and autoscaling.

You can use either of these modes of operation to run your GKEworkloads or to operate entire GKE clusters, as described in thefollowing sections.

Autopilot and Standard mode for workloads

Best Practice:

Use Autopilot for most workloads, unless your application requiresprivileges or configuration options that don't meet the Autopilotconstraints.

You can run your workloads in Autopilot or Standard, dependingon your use case and on the requirements of the workload. The followingtable describes the differences between Autopilot workloads andand Standard workloads:

Table 1. Comparison of Autopilot and Standard workloads
Autopilot workloadsStandard workloads
Ideal for most production workloads.Ideal for workloads that require special privileges, or for when you need granular control over your workload infrastructure.
GKE provisions compute resources based on your Kubernetes workload specifications.You manage resource allocation and the placement of Pods on specific nodes.
Autopilot workloads can use the Autopilotcontainer-optimized compute platform, which is a scalable, cost-efficient platform that works well for most general-purpose workloads.You must plan the capacity of nodes in Standard node pools based on your workload requirements. You must also configure settings for scaling your node sizes and quantity.
GKE maximizes node resource usage bybin-packing Pods onto nodes.You plan workload placement based on the size of the nodes that you create. To minimize wasted node resources, you must also manage effective bin-packing of Pods.
GKE applies many best practice configurations and constraints by default, such as preventing privileged Pods.You can configure Pods in any way that you want, such as by enabling privileged mode.
You can run Autopilot workloads in Autopilot clusters orin Standard clusters.You can run Standard workloads in only Standard clusters.

Workloads run in Autopilot or Standard mode depending onthe workload and cluster configuration, as follows:

About Autopilot workloads in Standard clusters

You can run workloads in Autopilot mode in your Standardclusters, which lets you have granular control over your cluster settings andnode pool infrastructure while also getting many of the automation, scaling, andpricing benefits of Autopilot for a subset of your applications.

Run Autopilot workloads in Standard clusters when youneed the flexibility of a Standard mode cluster, but you also haveworkloads that can benefit from the cost-efficiency and automation ofAutopilot mode.

To run Autopilot workloads in a Standard cluster, you use acompute class, which is a GKE feature that lets youdeclaratively control node configuration options during scaling.GKE manages bin-packing, resource scaling, and the underlyinginfrastructure for the Autopilot workloads that you run in yourStandard cluster.

For more information, seeAbout Autopilot workloads in GKE Standard.

Autopilot and Standard mode for clusters

To run workloads in GKE, you create a GKE cluster.Similarly to selecting a mode for your workloads, you can use Autopilotor Standard mode to operate your entire cluster. The mode that you usefor the cluster affects how much flexibility and control you have over yourinfrastructure and cluster settings, as follows:

Table 2. Comparison of cluster modes
Autopilot clustersStandard clusters
GKE manages the cluster infrastructure, including nodes and scaling. GKE configures various settings by default, such as security constraints.You control the cluster infrastructure and can change most settings. In Standard clusters, you can configure your ownnode pools, which are groups of nodes that share characteristics. You also configure node settings, security policies and constraints, and scaling.
Autopilot clusters are pre-configured with Google's best practices and run only Autopilot workloads.Standard clusters can run both Autopilot and Standard workloads.

For more information about the detailed differences between Autopilotand Standard clusters, seeCompare features in Autopilot and Standard clusters.

Unless you require the granular control and flexibility of Standardclusters, we recommend that you use Autopilot clusters. For aninteractive walkthrough that sets up an Autopilot cluster and creates ahello-world application, go to the Autopilot walkthrough in theGoogle Cloud console:

Go to walkthrough

Pricing

In GKE, the mode of operation that you use affects the chargesthat you incur. For more information, seeGoogle Kubernetes Engine pricing.

What's next

Except as otherwise noted, the content of this page is licensed under theCreative Commons Attribution 4.0 License, and code samples are licensed under theApache 2.0 License. For details, see theGoogle Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2026-02-18 UTC.