DirectML is Microsoft’s high-performance, hardware-accelerated DirectX 12 library for machine learning on Windows. The GitHub repository has collected around 2,600 stars and remains maintained, though Microsoft notes the project is now in maintenance mode. DirectML provides GPU acceleration for common machine learning operators across a broad range of hardware, covering DirectX 12-capable GPUs from AMD, Intel, NVIDIA, and Qualcomm through a single API.
Rather than writing separate CUDA, ROCm, or vendor-specific code, developers target DirectML once and let the DirectX 12 stack dispatch work to whatever GPU the user has installed. It underpins ML features across the Windows ecosystem: it serves as an execution provider for ONNX Runtime, powers Windows ML, and has accelerated everything from image upscaling in games to local AI features in creative apps. The repository contains samples, the machine learning API headers, and documentation for integrating the library.
Because DirectML rides on DirectX 12, it benefits automatically from the driver ecosystem Windows users already have, avoiding the installation friction of separate compute stacks. Microsoft positioned it as the GPU foundation beneath Windows ML and ONNX Runtime’s DirectML execution provider, which is how many desktop AI features reach machines with integrated or older discrete GPUs. Even in maintenance mode, it remains the pragmatic choice for shipping one ML build to the whole Windows installed base.
DirectML is free: the library is a Windows platform component and the repository is open under the MIT license. There are no paid tiers or usage fees.
Windows application and game developers who need GPU-accelerated ML inference that works across the widest possible range of user hardware without vendor lock-in.