Leaderboard

An overview of pathology foundation models performance across multiple benchmarks

(46/46 models)
Model
ViT-H/14 (632M)
🥇
1/15
🥈
2/21
🥇
1/22
5/25
🥈
2/5
4/11
🥈
2/13
🥈
2/22
🥈
2/21
4/15
10
ViT-g/14 (1.1B)
-
🥇
1/21
-
🥇
1/25
---
6/22
--
3
ViT-g/14 (1.1B)
4/15
🥉
3/21
🥉
3/22
🥈
2/25
5/5
🥇
1/11
5/13
8/22
8/21
🥉
3/15
10
ViT-H/14 (682M)
🥈
2/15
4/21
5/22
🥉
3/25
🥉
3/5
--
🥇
1/22
9/21
7/15
8
ViT-L/14 (307M)
-
5/21
--------
1
UNI
ViT-L/16 (307M)
5/15
6/21
🥈
2/22
11/25
-
🥉
3/11
6/13
5/22
14/21
9/15
9
ViT-L/14 (303M)
-
7/21
9/22
12/25
-----
10/15
4
ViT-g/14 (1.1B)
9/15
8/21
4/22
10/25
-
🥈
2/11
🥉
3/13
9/22
12/21
6/15
9
ViT-B/16 (86M)
-
9/21
12/22
16/25
--
🥇
1/13
16/22
7/21
🥈
2/15
7
ViT-L/14 (307M)
13/15
10/21
15/22
14/25
-
6/11
-
19/22
20/21
14/15
8
ViT-L/16 (307M)
-
11/21
-
13/25
---
17/22
🥉
3/21
-
4
ViT-B/16 (86M)
11/15
12/21
13/22
18/25
-
5/11
12/13
18/22
19/21
12/15
9
V-FFN (202M)
-
13/21
-
24/25
---
20/22
10/21
-
4
ViT-H/14 (632M)
-
14/21
7/22
4/25
-
9/11
9/13
12/22
6/21
-
7
Swin-T/4 (28M)
-
15/21
22/22
23/25
-
10/11
7/13
-
16/21
-
6
ViT-L/14 (307M)
-
16/21
--------
1
ViT-L/14 (307M)
8/15
18/21
10/22
9/25
--
8/13
11/22
-
15/15
7
PLIP
ViT-B/32 (88M)
-
19/21
19/22
---
13/13
22/22
-
11/15
5
ViT-L/16 (306M)
-
21/21
--------
1
ViT-g/14 (1.1B)
🥉
3/15
--
7/25
---
4/22
-
5/15
4
ViT-B/8 (86M)
6/15
-
14/22
15/25
----
13/21
8/15
5
ViT-L/14 (307M)
7/15
--
20/25
--
11/13
--
13/15
4
ViT-B/16 (86M)
10/15
--
19/25
---
13/22
--
3
ViT-S/8 (22M)
12/15
--
21/25
------
2
ViT-S/16 (22M)
14/15
--
22/25
---
15/22
--
3
ViT-S/16 (22M)
15/15
-
11/22
-----
17/21
-
3
ViT-S/16 (21M)
--
17/22
-----
11/21
-
2
ViT-B/32 (86M)
--
21/22
----
21/22
--
2
ViT-B/16 (85M)
--
16/22
-------
1
ViT-B/14 (86M)
--
6/22
6/25
---
🥉
3/22
5/21
🥇
1/15
5
Atlas 2
ViT-G/8 (2.0B)
----
🥇
1/5
---
🥇
1/21
-
2
Atlas
ViT-H/14 (632M)
----
4/5
---
4/21
-
2
ViT-B/14 (86M)
------
10/13
10/22
--
2
KEEP
ViT-L/16 (307M)
-------
7/22
--
1
ViT-g/14 (1.1B)
---
8/25
---
14/22
--
2
ViT-S/8 (22M)
---
17/25
--
4/13
---
2
ResNet-50 (25M)
-
20/21
-
25/25
-
11/11
----
3
CHIEF
Swin-T (CTransPath) (28M)
-
17/21
--------
1
Ciga
ResNet-18 (11.7M)
--------
21/21
-
1
RetCCL
ResNet-50 (23M)
--------
15/21
-
1
RudolfV
ViT-L/14 (307M)
--------
18/21
-
1
ViT-S/16 (22M)
-----
8/11
----
1
ViT-B/16 (85M)
-----
7/11
----
1
ViT-B/16 (86M)
--
20/22
-------
1
ViT-B/16 (86M)
--
18/22
-------
1
ViT-B/16 (86M)
--
8/22
-------
1

Direct comparison

Compare pathology foundation models on a common set of benchmarks

Robustness:
Performance:

Scroll to zoom, drag to pan. Double-click to reset.

12 models with data for all selected benchmarks.

Scaling laws

Explore how model size correlates with performance or robustness across benchmarks

Benchmarks:

Scroll to zoom, drag to pan. Double-click to reset.

Linear regression: R² = 0.408

15 models with data for all selected benchmarks.