Windows ML Runtime Intel OpenVINO Execution Provider 1.8.95.0

Windows ML Runtime Intel OpenVINO Execution Provider 2 integrates Intel’s OpenVINO inference engine with the Windows ML runtime to enable accelerated model execution on Intel hardware. It acts as a bridge between Windows-native ML workflows and OpenVINO’s optimization and device backend, targeting deployment scenarios on Windows devices with Intel CPUs or integrated GPUs.

Functionally, the provider offloads supported operators to OpenVINO, leveraging model optimizations, reduced-precision kernels, and device-specific drivers to improve latency and throughput for common vision and inference workloads. It integrates into the Windows ML execution pipeline via provider registration, exposes device selection and asynchronous execution options, and can be paired with existing ONNX/Windows ML model assets after appropriate conversion or validation. Diagnostics and logging hooks are available for performance tuning and troubleshooting.

• Purpose: hardware-accelerated inference for Windows applications using Intel compute backends.
• Features: operator offload to OpenVINO, support for reduced-precision inference (where applicable), asynchronous execution, and integration with Windows ML APIs.
• Usability: straightforward provider registration within the Windows ML runtime; requires installation and configuration of the matching OpenVINO runtime and Intel drivers.
• Compatibility and limits: best suited for Intel platforms; operator coverage depends on OpenVINO support and some models may require fallback to software execution or conversion steps.
• Developer experience: benefits from existing Windows tooling and runtime diagnostics, but version and driver alignment are important for stable results.