onnx

ONNX is an open format for representing machine learning models, enabling interoperability between different AI frameworks and hardware. On Windows, ONNX models can be deployed locally using Windows Machine Learning (Windows ML) and accelerated via DirectML on any DirectX 12 GPU. This allows developers to run AI inference on billions of Windows devices without relying on cloud services. Discussions cover the architectural shift of integrating ONNX with dedicated Vision Processing Units (VPUs) for on-device AI, as well as Microsoft's release of open-weight AI models that leverage ONNX for custom, privacy-first innovation. The tag explores how ONNX facilitates efficient AI deployment across Windows, Azure, and edge devices.