OpenVINO
Toolkit for deploying inference neural network model on Intel hardware
From Wikipedia, the free encyclopedia
OpenVINO is an open-source software toolkit developed by Intel for optimizing and deploying deep learning models. It supports several popular model formats[2] and categories, such as large language models, computer vision, and generative AI.
| OpenVINO | |
|---|---|
 | |
| Developer | Intel Corporation |
| Initial release | May 16, 2018 |
| Stable release | 2026.0
/ February 2026.[1] |
| Written in | C++ |
| Operating system | Cross-platform |
| License | Apache License 2.0 |
| Website | openvino |
| Repository | github |
| As of | February 2026 |
OpenVINO is optimized for Intel hardware, but offers support for ARM/ARM64 processors.[2] It sees great use[according to whom?] in AI Sound Processing drivers when tied with Intel's Gaussian & Neural Accelerator (GNA).
Based in C++, it extends API support for C and Python, as well as Node.js (in early preview).
OpenVINO is cross-platform and free for use under Apache License 2.0.[3]
Workflow
The simplest OpenVINO usage involves obtaining a model and running it as is. Yet for the best results, a more complete workflow is suggested:[4]
- obtain a model in one of supported frameworks,
- convert the model to OpenVINO IR using the OpenVINO Converter tool,
- optimize the model, using training-time or post-training options provided by OpenVINO's NNCF.
- execute inference, using OpenVINO Runtime by specifying one of several inference modes.
OpenVINO model format
OpenVINO IR[5] is the default format used to run inference. It is saved as a set of two files, *.bin and *.xml, containing weights and topology, respectively. It is obtained by converting a model from one of the supported frameworks, using the application's API or a dedicated converter.
Models of the supported formats may also be used for inference directly, without prior conversion to OpenVINO IR. Such an approach is more convenient but offers fewer optimization options and lower performance, since the conversion is performed automatically before inference. Some pre-converted models can be found in the Hugging Face repository.[6]
The supported model formats are:[7]
- PyTorch
- TensorFlow
- TensorFlow Lite
- ONNX (including formats that may be serialized to ONNX)
- PaddlePaddle
- JAX/Flax
