Skip to main content
A growing number of Windows machines now use win-arm64 processor architecture. This is a different processor architecture than win-64 processor architecture, which has been the standard platform architecture in most Windows PCs for decades. Because these two architectures use different instruction sets, software compiled for one is not compatible with the other. Anaconda now publishes win-arm64 packages built specifically for Arm processors on Windows. Because win-arm64 Windows supports both architectures with a built-in compatibility layer, you can run win-64 applications on a win-arm64 device with minimal configuration. This guide covers the pros and cons of your architecture choices, how to install packages for each processor type, and how to make changes to your environments when needed.

​
Choosing an architecture setup

If you have a Windows on Arm (win-arm64) device, you need to choose which Miniconda installer to use. You can install the win-arm64 version to match your hardware, install the win-64 version to keep full access to the existing package library, or install win-arm64 and create specific environments for win-64 packages as needed. The following table compares these options:
Anaconda currently only publishes win-arm64 installers for Miniconda. Anaconda Distribution installers are not yet available for win-arm64.

​
How to decide

Check whether the packages your projects are using are available for win-arm64. Run the following command for each package:
You can also browse the win-arm64 package listing on repo.anaconda.com.
win-arm64 environments require Python 3.12 or later.
  • If everything you need is available, install win-arm64 Miniconda and use win-arm64 packages for the best speed and performance.
  • If your projects are using packages that require win-64, you do not have to migrate to win-arm64. Install a win-64 (x86-64) version of Miniconda and use Anaconda’s full library of win-64 packages.
    The win-64 package architecture is not being deprecated. It remains a fully supported platform architecture with no end of service date. You are not required to migrate to win-arm64.
  • If some projects have packages that are available for win-arm64 and others do not, install the win-arm64 Miniconda and create --platform win-64 environments for the projects that need them.

​
Creating a win-64 environment on a win-arm64 installation

If you have installed win-arm64 Miniconda but need to create an environment that uses win-64 packages, use the --platform flag with conda create:
Conda records the target platform in the environment’s configuration, so any future conda install commands in that environment will continue to pull win-64 packages without needing to use the --platform flag again.
An environment’s platform can only be configured when it is created.

​
Migrating environments to win-arm64

To migrate your project environments to use win-arm64 packages, you must export your existing environments, uninstall your current Miniconda installation, reinstall with the win-arm64 installer, and recreate your environments.
1

Export your environments from the win-64 installation

Repeat this step for every environment you want to migrate.
For long-lasting reproducibility, export your environment to a conda lockfile.
2

Uninstall your current Miniconda installation

Follow the uninstallation instructions for Miniconda.
3

Install Miniconda for win-arm64

Download and install the win-arm64 Miniconda installer from anaconda.com/download. For architecture-specific instructions, see Installing Miniconda.
4

Recreate your environments on the new installation

Repeat this step for every environment you want to recreate.If environment install fails, see Windows Arm64 environment fails to recreate.