AIME
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
AIME24
86Accuracy
12
AIME
1.64Decoding Speedup
12
AIME '24
41.66Pass@1
12
AIME 24
53.2Accuracy
12
AIME 24
18.33Mean@16
12
AIME 25
18.96Mean@16
12
AIME24
34.18Avg@8 Accuracy
12
AIME25
13,040Reasoning Steps (Tokens)
12
AIME 24
18,923Thinking Tokens
12
AIME 2026
51.3Avg@32
12
AIME 25
5.83Avg@8
12
AIME 24
9.17Avg@8
12
AIME25
5.83Avg@8
12
AIME 24
9.17Avg@8
12
AIME 25
7.08Avg@8
12
AIME24
10.83Avg@8
12
AIME out-of-domain evaluation 2025
45.1Average Score @32
12
AIME25
33.3EM Accuracy
12
AIME24
36.7EM Accuracy
12
AIME 2025
66.7Accuracy
12
AIME 1983–2024
0Average False Positive Rate
12
AIME24 Full Standard
1.9Average Score @16
12
AIME24
77.7Accuracy
12
AIME24
47.43Wall-clock Time (s)
12
AIME 2024 (test)
28.9Accuracy
12