Runtimes and runtime templates

This page describes runtimes and runtime templates for Colab Enterprise.

A runtime is a Google-provisioned virtual machine (VM) that can run the code in your notebook (IPYNB file).

A runtime template is a VM configuration that specifies a machine type and other characteristics of the VM, as well as common settings such as the network and whether public internet access is enabled. When you create a runtime, its VM is created according to the specifications of a runtime template.

Workflow

You can create a Colab Enterprise notebook and run code in it without needing to understand runtimes. When you run your code for the first time, Colab Enterprise provisions a default runtime and runs your code on it. Colab Enterprise continues to use the default runtime until you connect to a different runtime.

To configure a runtime for specific needs, you must:

1. Create a runtime template with the configuration that you need.

2. Create a runtime based on that template.

3. Connect to the runtime from your notebook and run your code.

Runtimes

This section describes the characteristics of runtimes.

Default runtime

If you don't create a runtime, a runtime without accelerators is created for you the first time that you run code or connect to the default runtime. You can use the default runtime without accelerators, or switch to a default runtime that includes accelerators.

When Colab Enterprise creates a default runtime, it first creates a corresponding default runtime template. Both the runtime and runtime template appear in the Google Cloud console, where you can get information about them.

If you use default runtimes without accelerators, Colab Enterprise creates one default runtime without accelerators and its corresponding runtime template, per project, per region.

If you use default runtimes with accelerators, Colab Enterprise creates one default runtime with accelerators and its corresponding runtime template, per project, per region.

Default runtime specifications

Default runtimes and their corresponding default runtime templates include the following specifications:

Python versions

Starting in May, 2025, when Colab Enterprise creates a default runtime template, the default runtime template is configured to use the latest version of Python (currently Python 3.11). This means that when a new version of Python is available in Colab Enterprise, new runtimes created from the default runtime template will use the latest Python version.

If your default runtime templates were created in May 2025 or earlier, they might be configured to use Python 3.10. To update a default runtime template to use the latest Python version, you must delete the default runtime template. Then, when a new default runtime template is created, it will be configured to use the latest Python version.

To update the Python version of a runtime template that you created, see Change a runtime template's Python version.

The Python version of existing runtimes doesn't automatically change, even when a new Python version is available in Colab Enterprise. Changing the Python version of an existing runtime isn't supported. Instead, you must create a new runtime derived from a runtime template that uses the Python version that you want.

Change a runtime template's Python version

To change a runtime template's Python version, you can use the REST API.

Before using any of the request data, make the following replacements:

• REGION: the region where your runtime template is located.
• PROJECT_ID: your project ID.
• NOTEBOOK_RUNTIME_TEMPLATE_ID: the ID of your runtime template.
• RELEASE_NAME: the release name of the image that you want to use. For example, py311 is the release name for the Python 3.11 image. To set the image to Latest, leave RELEASE_NAME empty.

HTTP method and URL:

PATCH https://REGION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/REGION/notebookRuntimeTemplates/NOTEBOOK_RUNTIME_TEMPLATE_ID?updateMask=software_config.colab_image.release_name

Request JSON body:

{
  softwareConfig: {
    colabImage: {
      releaseName: "RELEASE_NAME"
    }
  }
}

To send your request, choose one of these options:

curl -X PATCH \
     -H "Authorization: Bearer $(gcloud auth print-access-token)" \
     -H "Content-Type: application/json; charset=utf-8" \
     -d @request.json \
     "https://REGION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/REGION/notebookRuntimeTemplates/NOTEBOOK_RUNTIME_TEMPLATE_ID?updateMask=software_config.colab_image.release_name"

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
    -Method PATCH `
    -Headers $headers `
    -ContentType: "application/json; charset=utf-8" `
    -InFile request.json `
    -Uri "https://REGION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/REGION/notebookRuntimeTemplates/NOTEBOOK_RUNTIME_TEMPLATE_ID?updateMask=software_config.colab_image.release_name" | Select-Object -Expand Content

For more information, see the notebookRuntimeTemplates.patch REST API documentation.

Runtimes are regional

Runtimes are regional, and must be located in the same region as your notebook.

Runtimes are exclusive to a user

Runtimes are exclusive to a user. You can't share your runtimes with others and you can't use their runtimes. Even if you share a notebook with someone, they don't have access to your runtime and must run the notebook on their own runtime.

However, users can create their own runtimes based on the same runtime template. See Runtime templates can be shared with users.

Managing your runtime's files

You can access and modify files on your runtime by using the  Files button in Colab Enterprise. You can also upload files to your runtime.

When your runtime is deleted, any files that you upload are deleted. If you've modified files, those files will revert back to their original state when a new runtime is created. See the auto deletion time period in the Default runtime specifications section.

Runtimes can be shared by notebooks

You can connect more than one notebook to the same runtime but this isn't recommended. Multiple notebooks sharing the same runtime can cause problems, including:

• The notebooks may run slowly because they are sharing the runtime's compute resources.

• The code in one notebook can affect the state of the runtime's VM for the other notebook. For example, if you install version 1.3 of a package in one notebook, but your other notebook requires version 1.2, you might have a version incompatibility when you run the second notebook.

Instead of running code from multiple notebooks on the same runtime, you can create a runtime for each notebook based on the same runtime template. See You can create multiple runtimes from the same runtime template.

Runtime charges

You are charged for runtime processing based on the machine type and accelerators that you use. You are also charged for boot disk and data disk storage. Boot disks and data disks are deleted when the runtimes are deleted.

Runtimes always include a 100 GiB SSD persistent disk in addition to the data disk specified in the runtime template. The boot disk isn't configurable.

While runtimes are shut down:

• Compute charges aren't generated because there isn't any CPU or GPU usage.
• Boot disk and data disk storage is charged.

For more information, see Colab Enterprise pricing.

You can create multiple runtimes from the same runtime template

You can create multiple runtimes from a single non-default runtime template. This way, you can run multiple notebooks on VMs with the same configuration without sharing the resources of that VM.

Runtime templates

This section describes the characteristics of runtime templates.

Runtime templates are regional

Runtime templates are regional. When you create a runtime, it is in the same region as the runtime template, and the runtime can only be used with notebooks in that region.

Runtime templates can be shared with users

You can share a runtime template with other users. This way, each user can create their own runtime based on the same runtime template, and therefore multiple users can run the same notebook on the same VM configuration.

To learn more, see Manage access to a runtime template.

Configuration

Runtime templates determine the configuration of your runtime's VM. You can specify any of the following:

Region

Choose the region where your notebook is. Runtimes created from runtime templates are in the same region as the runtime template. For the notebook to run code in the runtime, the notebook and runtime must be in the same region.

Machine type

You can configure the machine type of your runtime template. For information on machine types, see the Machine families resource and comparison guide.

Data disk

You can configure the disk type and disk size of the data disk in your runtime template.

When your runtime is created, it also includes a boot disk. The boot disk is a 100 GiB SSD persistent disk and can't be configured.

Accelerators

You can add accelerators to your runtime template configuration. Accelerator availability is based on the machine type and region.

Network

Specify your network or use the project's default Virtual Private Cloud network.

Idle shutdown

Colab Enterprise enables idle shutdown by default. This setting shuts down your instance after 180 minutes of inactivity. You can change the number of minutes of inactivity before shutdown, or you can disable idle shutdown.

Public internet access

To get more control over the security of your notebook and other resources, you can turn off public internet access. If you configure your runtime template this way, notebooks that use runtimes created from the runtime template won't have access to the public internet.

End-user credential authentication

By default runtime templates use end-user credential authentication to give your notebook access to Application Default Credentials (ADC). You can use this access to ADC to run code in your notebook that interacts with Vertex AI and other Google Cloud services. When end-user credentials are enabled, your notebook uses the credentials of whoever runs the notebook to determine access to Google Cloud resources.

If end-user credentials aren't enabled, you can get access to ADC by authenticating the user through your notebook's code. To learn more, see the "User credentials" section of Set up ADC for a local development environment.

Runtime template charges

There are no charges for creating or storing runtime templates. See also Runtime charges.

What's next

• Create a runtime based on a runtime template.
• Connect to a runtime.
• Create a runtime template.