Benchmark Report 2025

Visual Product
Search Benchmark

A comprehensive evaluation of state-of-the-art embedding models for visual product search. Comparing 10 models across 8 diverse datasets spanning automotive parts, furniture, hardware, and industrial components.

8
Datasets
10
Models
70+
Evaluations
5
Metrics

Explore the Benchmark

Navigate through different aspects of our evaluation study.

Key Findings

Highlights from our benchmark evaluation

Top Performer
nyris General V5.1

Achieves highest average Precision@1 across all datasets

Best on Stanford OP
86.82%

Precision@1 achieved by nyris General V5.1

Hardest Dataset
ARaymond

Industrial parts with highest variance in model performance