AIME
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
AIME Art of Problem Solving 2025 2026
100Score
3
AIME 2024
53.33Pass@32
3
AIME 2025 (test)
50Pass@1 Rate
3
AIME24
93.33Exact Match (EM)
3
AIME All Non-Easy
22.2Recovery Rate
3
AIME Med. + Hard
10.8Recovery Rate
3
AIME OOD
77.92CPT 85%
3
AIME
84.88CPT (90%)
3
AIME OOD
71.82CPT (80%)
3
AIME Streaming Answers
0.032Brier Score
3
AIME
0.075ECE
3
AIME 2025
47.8Pass@64
3
AIME 2024
62.1Pass@64
3
AIME 1983-2024 (val)
81Accuracy
3
AIME 2025
19.16Avg@32 Accuracy
3
AIME 2024
31.25AIME 2024 Avg@32 Accuracy
3
AIME Out-of-distribution 2024 (test)
14.8Accuracy
3
AIME OpenR1-Math Harder Subset 2025
66.1Accuracy
3
AIME OpenR1-Math Harder 2024
72.7Accuracy
3
AIME25
19.1Average Accuracy
3
AIME24
23.5Average Accuracy
3
AIME 24
2,768.5TPS
3
AIME25
2,979.8Throughput (TPS)
3
AIME 25
78.67Pass@1
3
AIME 24
83.3Pass@1
3