Specific Deep Learning VM Images images are available to suit your choice of framework and processor. There are currently images supporting TensorFlow, PyTorch, and generic high-performance computing, with versions for both CPU-only and GPU-enabled workflows. To find the image that you want, see the table below.
Deciding on an image family
Choose a Deep Learning VM image family based on the framework
and processor that you need.
The following table lists the most recent versions of image families,
organized by framework type.
To get the most recent version of an image, create an instance
by referencing an image family with latest in the name.
If you need a specific framework version, skip to Supported framework
versions.
| Framework | Processor | Image family name(s) |
|---|---|---|
| Base | GPU |
common-cu123common-cu122common-cu121common-cu118common-cu113common-cu110
|
| CPU | common-cpu |
|
| TensorFlow Enterprise | GPU | tf-ent-latest-gpu |
| CPU | tf-ent-latest-cpu |
|
| PyTorch | GPU | pytorch-latest-gpu |
| CPU | pytorch-latest-cpu |
Choosing an operating system
For most frameworks, Debian 11 is the default OS. Ubuntu 22.04
images are available for some frameworks.
They are denoted by the -ubuntu-2204
suffixes in the image family name (see Listing all available
versions).
Debian 10 and Debian 9 images have been deprecated.
PyTorch and TensorFlow Enterprise image families support A100 GPU accelerators.
TensorFlow Enterprise images
TensorFlow Enterprise image families provide a Google Cloud optimized distribution of TensorFlow. For more information about TensorFlow Enterprise including which versions are supported, see TensorFlow Enterprise overview.
Experimental images
Some Deep Learning VM image families are experimental, as indicated by the table of image families. Experimental images are supported on a best-effort basis, and may not receive refreshes on each new release of the framework.
Specifying an image version
You can reuse the same image even if the latest image is newer. This can be useful, for instance, if you are trying to create a cluster and you want to ensure that any images that are used to create new instances are always the same. You should not use the name of the image family in this situation because, if the latest image is updated, you'll have different images on some instances in your cluster.
Instead, you can determine what the exact name of the image is, incorporate the version number, and then use that specific image to spawn new instances in your cluster.
To find out the exact name of the latest image, use the following command in the Google Cloud CLI with your preferred terminal or in Cloud Shell. Replace IMAGE_FAMILY with the image family name for which you want to find out the latest version number.
gcloud compute images describe-from-family IMAGE_FAMILY \
--project deeplearning-platform-release
Look for the name field in the output and use the image name given there
when creating new instances.
Supported framework versions
Deep Learning VM supports each framework version based on a schedule to minimize security vulnerabilities. Review the Deep Learning VM framework support policy to understand the implications of the end-of-support and end-of-availability dates.
If you need a specific framework or CUDA version, see the following tables. To
find a specific VERSION_DATE for an image, see Listing
the available versions.
Base versions
| ML framework version | Current patch version | Supported accelerators | End of patch and support date | End of availability date | Image family name |
|---|---|---|---|---|---|
| Base-CPU (Python 3.10 / Debian 11) | Not applicable (N/A) | CPU only | Jul 1, 2024 | Jul 1, 2025 | common-cpu-VERSION_DATE-debian-11 |
| Base-cu123 (Python 3.10) | CUDA 12.3 | GPU (CUDA 12.3) | Oct 19, 2024 | Oct 19, 2025 | common-cu123-VERSION_DATE-debian-11-py310 |
| Base-cu122 (Python 3.10) | CUDA 12.2 | GPU (CUDA 12.2) | June 28, 2024 | June 28, 2025 | common-cu122-VERSION_DATE-debian-11-py310 |
| Base-cu121 (Python 3.10) | CUDA 12.1 | GPU (CUDA 12.1) | Feb 28, 2024 | Feb 28, 2025 | common-cu121-VERSION_DATE-debian-11-py310 |
| Base-cu118 (Python 3.10) | CUDA 11.8 | GPU (CUDA 11.8) | Jul 1, 2024 | Jul 1, 2025 | common-cu118-VERSION_DATE-debian-11-py310 |
| Base-cu113 (Python 3.10) | CUDA 11.3 | GPU (CUDA 11.3) | Jan 1, 2024 | Jan 1, 2025 | common-cu113-VERSION_DATE-debian-11-py310 |
| Base-cu113 (Python 3.7) | CUDA 11.3 | GPU (CUDA 11.3) | Sep 1, 2023 | Sep 1, 2024 | common-cu113-VERSION_DATE-py37 |
| Base-cu110 (Python 3.7) | CUDA 11.0 | GPU (CUDA 11.0) | Sep 1, 2023 | Sep 1, 2024 | common-cu110-VERSION_DATE-py37 |
| Base-CPU (Python 3.7) | Not applicable (N/A) | CPU only | Sep 1, 2023 | Sep 1, 2024 | common-cpu-VERSION_DATE-debian-10 |
TensorFlow versions
| ML framework version | Current patch version | Supported accelerators | End of patch and support date | End of availability date | Image family name |
|---|---|---|---|---|---|
| 2.17 (Python 3.10) | 2.17.0 | CPU only | Jul 11, 2025 | Jul 11, 2026 | tf-2-17-cpu-VERSION_DATE-py310 |
| 2.17 (Python 3.10) | 2.17.0 | GPU (CUDA 12.3) | Jul 11, 2025 | Ju1 11, 2026 | tf-2-17-cu123-VERSION_DATE-py310 |
| 2.16 (Python 3.10) | 2.16.2 | CPU only | Jun 28, 2025 | Jun 28, 2026 | tf-2-16-cpu-VERSION_DATE-py310 |
| 2.16 (Python 3.10) | 2.16.2 | GPU (CUDA 12.3) | Jun 28, 2025 | Jun 28, 2026 | tf-2-16-cu123-VERSION_DATE-py310 |
| 2.15 (Python 3.10) | 2.15.0 | CPU only | Nov 14, 2024 | Nov 14, 2025 | tf-2-15-cpu-VERSION_DATE-py310 |
| 2.15 (Python 3.10) | 2.15.0 | GPU (CUDA 12.2) | Nov 14, 2024 | Nov 14, 2025 | tf-2-15-cu122-VERSION_DATE-py310 |
| 2.14 (Python 3.10) | 2.14.0 | CPU only | Sep 26, 2024 | Sep 26, 2025 | tf-2-14-cpu-VERSION_DATE-py310 |
| 2.14 (Python 3.10) | 2.14.0 | GPU (CUDA 11.8) | Sep 26, 2024 | Sep 26, 2025 | tf-2-14-cu118-VERSION_DATE-py310 |
| 2.13 (Python 3.10) | 2.13.0 | CPU only | Jul 5, 2024 | Jul 5, 2025 | tf-2-13-cpu-VERSION_DATE-py310 |
| 2.13 (Python 3.10) | 2.13.0 | GPU (CUDA 11.8) | Jul 5, 2024 | Jul 5, 2025 | tf-2-13-cu118-VERSION_DATE-py310 |
| 2.12 (Python 3.10) | 2.12.0 | CPU only | June 30, 2024 | June 30, 2025 | tf-2-12-cpu-VERSION_DATE-py310 |
| 2.12 (Python 3.10) | 2.12.0 | GPU (CUDA 11.8) | June 30, 2024 | June 30, 2025 | tf-2-12-cu113-VERSION_DATE-py310 |
| 2.11 (Python 3.10) | 2.11.0 | CPU only | Nov 15, 2022 | Nov 15, 2023 | tf-2-11-cpu-VERSION_DATE-py310 |
| 2.11 (Python 3.10) | 2.11.0 | GPU (CUDA 11.3) | Nov 15, 2022 | Nov 15, 2023 | tf-2-11-cu113-VERSION_DATE-py310 |
| 2.11 | 2.11.0 | CPU only | Nov 15, 2023 | Nov 15, 2024 | tf-2-11-cpu-VERSION_DATE-py37 |
| 2.11 | 2.11.0 | GPU (CUDA 11.3) | Nov 15, 2023 | Nov 15, 2024 | tf-2-11-cu113-VERSION_DATE-py37 |
| 2.10 | 2.10.1 | CPU only | Nov 15, 2023 | Nov 15, 2024 | tf-2-10-cpu-VERSION_DATE-py37 |
| 2.10 | 2.10.1 | GPU (CUDA 11.3) | Nov 15, 2023 | Nov 15, 2024 | tf-2-10-cu113-VERSION_DATE-py37 |
| 2.9 | 2.9.3 | CPU only | Nov 15, 2023 | Nov 15, 2024 | tf-2-9-cpu-VERSION_DATE-py37 |
| 2.9 | 2.9.3 | GPU (CUDA 11.3) | Nov 15, 2023 | Nov 15, 2024 | tf-2-9-cu113-VERSION_DATE-py37 |
| 2.8 | 2.8.4 | CPU only | Nov 15, 2023 | Nov 15, 2024 | tf-2-8-cpu-VERSION_DATE-py37 |
| 2.8 | 2.8.4 | GPU (CUDA 11.3) | Nov 15, 2023 | Nov 15, 2024 | tf-2-8-cu113-VERSION_DATE-py37 |
| 2.6 (py39) | 2.6.5 | CPU only | Sep 1, 2023 | Sep 1, 2024 | tf-2-6-cpu-VERSION_DATE-py39 |
| 2.6 (py39) | 2.6.5 | GPU (CUDA 11.3) | Sep 1, 2023 | Sep 1, 2024 | tf-2-6-cu110-VERSION_DATE-py39 |
| 2.6 (py37) | 2.6.5 | CPU only | Sep 1, 2023 | Sep 1, 2024 | tf-2-6-cpu-VERSION_DATE-py37 |
| 2.6 (py37) | 2.6.5 | GPU (CUDA 11.3) | Sep 1, 2023 | Sep 1, 2024 | tf-2-6-cu110-VERSION_DATE-py37 |
| 2.3 | 2.3.4 | CPU only | Sep 1, 2023 | Sep 1, 2024 | tf-2-3-cpu |
| 2.3 | 2.3.4 | GPU (CUDA 11.3) | Sep 1, 2023 | Sep 1, 2024 | tf-2-3-cu110-VERSION_DATE |
PyTorch versions
| ML framework version | Current patch version | Supported accelerators | End of patch and support date | End of availability date | Image family name |
|---|---|---|---|---|---|
| 2.2 (Python 3.10) | 2.2.0 | CUDA 12.1 | Jan 30, 2025 | Jan 30, 2026 | pytorch-2-2-VERSION_DATE-py310 |
| 2.1 (Python 3.10) | 2.1.0 | CUDA 12.1 | Oct 4, 2024 | Oct 4, 2025 | pytorch-2-1-VERSION_DATE-py310 |
| 2.0 (Python 3.10) | 2.0.0 | CUDA 11.8 | Mar 15, 2024 | Mar 15, 2025 | pytorch-2-0-VERSION_DATE-py310 |
| 1.13 (Python 3.10) | 1.13.1 | CUDA 11.3 | Dec 8, 2023 | Dec 8, 2024 | pytorch-1-13-VERSION_DATE-py310 |
| 1.13 | 1.13.1 | CUDA 11.3 | Dec 8, 2023 | Dec 8, 2024 | pytorch-1-13-VERSION_DATE-py37 |
| 1.12 | 1.12.1 | CUDA 11.3 | Sep 1, 2023 | Sep 1, 2024 | pytorch-1-12-VERSION_DATE-py310 |
List all available versions using gcloud CLI
You can also list all available Deep Learning VM images using the following gcloud CLI command:
gcloud compute images list \
--project deeplearning-platform-release \
--format="value(NAME)" \
--no-standard-images
Image families are named in the format
FRAMEWORK-VERSION-CUDA_VERSION(-experimental),
where FRAMEWORK is the target library,
VERSION is the framework version, and
CUDA_VERSION is the version of the CUDA stack, if present.
For example, an image from the family
tf-ent-2-13-cu113 has
TensorFlow Enterprise 2.13 and CUDA 11.3.
What's next
Create a new Deep Learning VM instance using the Cloud Marketplace or using the command line.