An open format defining a common set of operators and a file format to represent machine learning models for interoperability across frameworks and runtimes.

Runtimes designed to execute ONNX models efficiently, enabling hardware-optimized inferencing across different accelerators and platforms.