Model Accuracy#

The following two tables present the absolute accuracy drop calculated as the accuracy difference between OV-accuracy and the original frame work accuracy for FP32, and the same for INT8, BF16 and FP16 representations of a model on three platform architectures. The third table presents the GenAI model accuracies as absolute accuracy values. Please also refer to notes below the table for more information.

A - Intel® Core™ i9-9000K (AVX2), INT8 and FP32
B - Intel® Xeon® 6338, (VNNI), INT8 and FP32
C - Intel® Xeon 8580 (VNNI, AMX), INT8, BF16, FP32
D - Intel® Flex-170, INT8 and FP16

Model Accuracy for INT8#
OpenVINO™ Model name	dataset	Metric Name	A, INT8	B, INT8	C, INT8	D, INT8
bert-base-cased	SST-2_bert_cased_padded	spearman@cosine	3.06%	2.89%	2.71%	2.71%
efficientdet-d0	COCO2017_detection_91cl	coco_precision		-0.59%		-0.55%
mask_rcnn_resnet50_atrous_coco	COCO2017_detection_91cl_bkgr	coco_orig_precision	-0.10%	-0.04%		-0.01%
mobilenet-v2	ImageNet2012	accuracy @ top1		-0.97%	-0.98%	-0.95%
resnet-50	ImageNet2012	accuracy @ top1		0.97%	0.94%	0.95%
ssd-resnet34-1200	COCO2017_detection_80cl_bkgr	map	-0.06%	-0.08%	-0.07%	-0.06%
ssd-mobilenet-v1-coco	COCO2017_detection_80cl_bkgr	coco-precision		-0.28%		-0.26%
yolo_v8n	COCO2017_detection_80cl	map	-0.11%	-0.05%

Model Accuracy for BF16, FP32 and FP16 (FP16: Flex-170 only. BF16: Xeon(R) 8480+ only)#
OpenVINO™ Model name	dataset	Metric Name	A, FP32	B, FP32	C, FP32	C, BF16	D, FP16
bert-base-cased	SST-2_bert_cased_padded	spearman@cosine	0.00%	0.00%	0.00%	-0.01%	0.02%
efficientdet-d0	COCO2017_detection_91cl	coco_precision		0.01%	0.00%	0.01%	0.00%
mask_rcnn_resnet50_atrous_coco	COCO2017_detection_91cl_bkgr	coco_orig_precision		-0.01%	-0.01%	0.05%	0.00%
mobilenet-v2	ImageNet2012	accuracy @ top1	0.00%	0.00%	0.00%	-0.18%	0.02%
resnet-50	ImageNet2012	accuracy @ top1	0.00%	0.00%	0.00%	0.01%	0.01%
ssd-resnet34-1200	COCO2017_detection_80cl_bkgr	map	0.02%	0.02%	0.01%	0.02%	0.02%
ssd-mobilenet-v1-coco	COCO2017_detection_80cl_bkgr	coco-precision	0.04%	0.01%	0.04%	0.08%	0.01%
yolo_v8n	COCO2017_detection_80cl	map	0.00%	0.00%	0.01%	0.05%	0.00%

Model Accuracy for VNNI-FP16, VNNI-INT4, AMX-FP16 and MTL-INT4 (Core Ultra iGPU)#
OpenVINO™ Model name	dataset	Metric Name	A, VNNI-FP16	B, VNNI-INT4	C, FAMX-FP16
chatGLM4	Wikiset	ppl
Gemma-2-9B	Wikitext	ppl		1.57	1.57
Llama-2-7b-chat	Wikiset	ppl		1.59	1.59
Llama-3-8b	Wikiset	ppl	1.45	1.48	1.45
Llama-3.2-3b-instruct	Wikiset	ppl	1.60	1.62	1.62
Mistral-7b	Wikitext	ppl	1.48	1.49	1.48
Phi3-mini-4k-instruct	Wikitext	ppl	1.55	1.55	1.55
Qwen-2-7B	Wikitext	ppl	1.52	1.53	1.52

Notes: For all accuracy metrics a “-”, (minus sign), indicates an accuracy drop. For perplexity (ppl) the values do not indicate a deviation from a reference but are the actual measured accuracy for the model.