The close partnership between GitHub and Arm has been focused on improving developer experience for anyone developing on Arm, with enhanced workflows that combine, native Arm runners, images with essential tools and libraries, and GitHub Copilot Extensions to accelerate development with AI assistance. Last month, we were excited to see GitHub’s release of Windows Runners for GitHub Actions in public preview. Combined with Arm enablement accelerated AI frameworks like ONNX on Windows and Arm native Pytorch builds for Windows announced a few weeks ago, it is easier than ever to develop, test and deploy AI applications for Windows on Arm.
Enterprises are increasingly adopting Windows Arm-based devices for their energy efficiency and mobility. However, the absence of native CI/CD pipelines on WoA has been a bottleneck. Native runners now enable seamless integration with enterprise workflows, allowing teams to build and test software in environments that mirror their production setups – with the most cost-effective runners available from GitHub!
The open-source community has long sought better support for WoA, especially for cross-platform tools and packages. Native GitHub Arm runners – available free in public repositories - lower the barrier to entry for contributors and maintainers aiming to support WoA without resorting to complex toolchains or emulators. The CPython project was one of the first open-source projects to add support for the WoA runners.
With the growing traction of Arm-powered Windows laptops, developers require robust pipelines to support applications on Windows client devices. These runners bring parity with x64 workflows, enabling the native building of Windows GUI apps, UWP, and .NET applications. The Arm-hosted runners are powered by Microsoft Azure Cobalt 100 processors, based on the Arm Neoverse N2, with 4 vCPUs and Armv9-A features, including Scalable Vector Extension 2 (SVE2).
The following output shows the result of running the wmic command on the Arm-hosted runner.
Backed by high-speed, SVE-enabled Arm servers in the cloud, these runners offer enterprise-grade performance for WoA workflows. This infrastructure meets the scale, and efficiency demands of modern CI/CD processes.
The integration of native Arm runners with GitHub Actions facilitates:
The GitHub-hosted windows-11-arm image includes a broad set of development tools, such as:
For the full list, see the GitHub partner images repo. You can use this repo to report issues, and request for more software packages to be added.
To use the Arm64 runner in your workflow, set:
runs-on: windows-11-arm
️Note: This label is currently available for public repositories only.
To further empower developers, the Arm extension for GitHub Copilot helps simplify and accelerate migration of applications to Windows on Arm. This tool acts as a smart assistant, analyzing your codebase and providing context-aware suggestions for making your code Arm-compatible.
Here are some example prompts to spark ideas as you build and deploy your applications:
@arm Can I use native GitHub hosted runners for building and testing my Windows Applications on Arm?
@arm How do I build my Qt application natively on Windows on Arm?
The extension reduces the friction of migration by offering real-time fixes and architecture-aware code suggestions.
With native GitHub-hosted Windows on Arm runners and AI-assisted tooling like the Arm Copilot extension, the future of Arm-native development is here:
Start integrating these tools today and join the movement toward a modern, high-performance, Arm-native software development lifecycle.
The new native PyTorch support, announced by Microsoft, means developers can now train and run models for a variety of use cases like directly on Arm64 Windows devices without relying on x64 emulation or cross-compilation. Learn how to install on your Windows machine today - PyTorch for Windows on Arm | Arm Learning Paths
With the runners, ML developers can now run automated testing and deployment workflows directly on Arm64 Windows infrastructure, crucial for validating models and pipelines in Arm-native environments.
Together, these advancements in the software ecosystem close the loop for end-to-end MLOps on Windows on Arm:
These advancements support a wide range of machine learning use cases, from generative AI models like Stable Diffusion, to natural language processing (NLP) tasks, to traditional classification and regression problems, making Windows on Arm a powerful, production-ready platform for modern ML workflows.
As hundreds of projects have adopted the Windows runners since they launched, developers are finding significant improvements and time savings in their workflows. Jeremy Sinclair, a Software & AppSec Architect and Microsoft MVP in Windows Development and .NET recently migrated a runner to the new Windows Arm runner and saw significant benefits.
“I was able to seamlessly configure and incorporate a runner for windows-11-arm into my pipeline. I've had to run it 3-4 times just to make sure I saw the results correctly, but the build running on the new Windows 11 Arm64 runner was 4x faster than my original Windows runner. I'm truly amazed at how little I had to change for the runner to work. Bravo Arm and GitHub teams!” -Jeremy Sinclair
Delighting developers is what the Arm & GitHub partnership is all about!
Get started today with the latest Learning Path to Automate Windows on Arm Builds with GitHub Arm-hosted Runners. If you are attending Microsoft Build this week – check out this online session to learn how to run PyTorch natively on Windows on Arm using GitHub Runners.
Stay tuned to learn more about how developers are using these new Windows runners to speed up builds and streamline their Windows on Arm CI/CD and MLOps pipelines.