Models

Every model we offer, each with the same card: a real result, what it finds, how it was measured and how to call it.

Our base models

80 classes · mAP50-95 0.711

80 classes · mAP50-95 0.603

1,000 classes · top-1 88.0%

Across the RF100-VL benchmark (100 datasets): mAP50-95 0.697

32 classes · mAP50-95 0.779

6 classes · mAP50-95 0.766

5 classes · mAP50-95 0.728

7 classes · mAP50-95 0.647

1 classes · mAP50-95 0.735