Table 5:

Model size and performance (bold-face denotes highest performance on each dataset).

Model	No. of params.	NCT	HMU	STARC-9
CNN	3.9 M	64.21	68.1	97.81
KimiaNet	7M	68.69	86.67	98.72
PathDino	22 M	77.35	89.64	98.37
ResNet50	24 M	62.59	85.71	98.64
EfficientNet-B7	64 M	82.47	84.45	98.8
ViT-base	86 M	84.25	90.29	98.09
HiPT	86 M	74.51	91.99	98.32
Histo-ViT	86 M	69.48	67.16	96.32
VIM4PATH	86 M	75.41	91.5	98.29
DeiT-B	86 M	81.63	90.05	98.65
Swin Trans-base	87 M	79.05	91.88	98.79
CTransPath	87 M	79.05	91.88	99
CONCH	87 M	75.69	90.02	98.19
ConvNeXT-base	88 M	82.82	91.07	98.36
UNI	88 M	80.43	91.8	98.26
ProvGigPath	305 M	74.18	90.6	98.37
Virchow	305 M	79.02	91.23	98.28