openai / mle-bench
MLE-bench is a benchmark for measuring how well AI agents perform at machine learning engineering
GitHub Repo 
15K
lines of main code
382 files
0.3K
lines of test code
7 files
15K
lines of other code
527 files
2y
age
624 days
30%
main code touched
1 year (14K LOC)
30%
new main code
1 year (14K LOC)
12K
py
2.3K
yaml
MJS
0.03K
mjs
30 days 90 days all time 2026 2025 2024
+0
-0
+0
-0
+394
-5
 
 
+390
-1
+4
-4
0 0 6

0

3

3

0 0 4

0

2

2

0 0 3

0

2

2

2026 2025 2024

generated by sokrates.dev (configuration) on 2026-06-24