AI Models verified for OpenVINO™#

The following is a list of models that have been verified to work with OpenVINO. Note that other models from OpenVINO-supported frameworks may also work properly but have not been tested.

AI Models that run on Intel® Core Ultra™ Processors with OpenVINO™ toolkit:

Search:

Category	topology	source framework	precision	infered with	CPU	GPU	NPU
3D Object Classification	pointnet-chair-100	onnx	FP32	ONNX Runtime OVEP	+	+	+
3D Object Classification	pointnet-chair-100	onnx	FP16	ONNX Runtime OVEP	+	+	+
Action Recognition	i3d-flow	tf	FP16	OpenVINO	+	+
Action Recognition	i3d-flow	tf	FP16-INT8	OpenVINO	+
Action Recognition	i3d-flow	tf	FP32	OpenVINO	+	+
Action Recognition	i3d-rgb	tf	FP16	OpenVINO	+
Action Recognition	i3d-rgb	tf	FP16-INT8	OpenVINO	+
Action Recognition	mc3_18	pytorch	FP32	torch.compile	+	+
Audio	data2vec2-base	onnx	FP32	ONNX Runtime OVEP	+	+	+
Audio	data2vec2-base	onnx	FP16	ONNX Runtime OVEP	+	+	+

Showing 1 to 10 of 1,791 entries

…

Marked cells indicate models that passed inference with no errors.

In the precision column, the “optimum-intel default” label corresponds to FP32 for small models
and INT8 for models greater than 1B parameters.
The results as of February 25 2025, for OpenVINO version 2025.0.
The models come from different public model repositories, such as Pytorch Model Zoo and
HuggingFace; they were executed on the designated hardware with OpenVINO either natively or
as a backend.