Samples

Windows ML Samples

Windows ML Runtime, Task API, and Server samples

The Windows ML Runtime API gives Windows applications one programming model for local AI on CPUs, GPUs, and NPUs. The same Runtime objects run ONNX models through ONNX Runtime and GGUF models through llama.cpp.

• Runtime: C++ and Python samples organized by scenario, covering getting started, vision, speech, language, model compilation, and ONNX Runtime interoperability.
• Tasks: higher-level Task API samples for text generation, chat completion, automatic speech recognition, and task composition over caller-owned Runtime objects.
• Server: host the Windows ML Server, which serves a language model to coding agents and other OpenAI-compatible clients on the same computer, with C++, C#, and Python clients.

Note

The Runtime, Task, and Server APIs are new and still evolving, and they may change. Report issues and share feedback through GitHub Issues.

Windows App SDK samples

Additional Windows ML samples for the Windows App SDK can be found in the WindowsAppSDK-Samples repository, alongside the rest of the Windows App SDK samples:

➡️ microsoft/WindowsAppSDK-Samples — Samples/WindowsML

Quick links

C++ (MSBuild)

• CppConsoleDesktop
• CppConsoleDesktop.FrameworkDependent
• CppConsoleDesktop.SelfContained
• CppConsoleDesktop.GenAI
• CppConsoleDll
• CppResnetBuildDemo

C++ (CMake)

• ResNetConsoleDesktop
• ResNetConsoleDesktop.SelfContained
• WinMLEpCatalog

C++ ABI

• CppAbiEPEnumerationSample

C# (.NET)

• CSharpConsoleDesktop
• ResnetBuildDemoCS
• HelloPhi
• cs-wpf
• cs-winforms
• cs-winui

Python

• SqueezeNetPython

Diagnostics

• capture-logs