9 Models

Embedding Models

State-of-the-art visual embedding models from leading providers, evaluated for product search performance.

Total Models

Open Source

Multi-Modal

Providers

Generic Models

General-purpose embedding models evaluated across all benchmark datasets.

Rank	Model	Accessibility	Modality	Embed Dim	Input Res	Avg P@1	Avg mAP@10
1	General V5.1	Proprietary	Vision	768	336	57.63%	50.34%
2	PE-Core L/14	Open Source	Vision	1024	336	42.87%	34.04%
3	Vertex AI Multi-Modal	Proprietary	Multi-Modal	1408	N/A	42.81%	34.81%
4	SigLIP2 SO400M	Open Source	Vision	1152	384	41.16%	31.22%
5	DINOv3 ViT-L/16	Open Source	Vision	1024	224	40.23%	29.80%
6	DINOv2 Large	Open Source	Vision	1024	224	34.01%	23.30%
7	Cohere Embed V4	Proprietary	Multi-Modal	1024	N/A	33.87%	28.55%
8	Jina Embeddings V4	Open Source	Multi-Modal	768	Dynamic	26.27%	18.08%
9	Nomic Embed MM 3B	Open Source	Multi-Modal	768	Dynamic	25.78%	17.28%

Specialized models trained for specific product domains, evaluated only on their target datasets.

Model	Accessibility	Target Domain	Embed Dim	Input Res	P@1	mAP@10
Automotive V1	Proprietary	Intercars	768	336	32.49%	33.96%

Open Source

Freely available model weights that can be deployed on your own infrastructure. Offers flexibility and control over the inference pipeline.

Proprietary

Closed-source models accessed via API or with restricted access. Often optimized for specific use cases with specialized training data.

N/A

Input resolution is not available from the model's official documentation or specifications.

Dynamic

The input size is dynamically adjusted based on the model's resizing scheme and image resolution.