AIME
Benchmarks
TaskDatasetSOTA resultTrendResults
AIME 2025 edition (test)
4,377Optimization Tokens
3
AIME (test)
3,098Token Usage (Optimization Phase, Thousands)
3
AIME 2025
13.33Pass@1
3
AIME 24
3.63Speedup
2
AIME-90 (test)
32.2Accuracy (Full)
2
AIME-25
0.886Sel-AUC
2
AIME24
73.67Pass@2
2
AIME 2026 (test)
14Baseline Score
2
AIME 2025
0.63Pass@1
2
AIME 2024
9.27Pass@1
2
AIME 2025
6.7Turn 1 Score
2
AIME 2024
6.7Turn 1 Score
2
AIME 2025
0.18Token Efficiency Ratio
2
AIME 2025
45.6Avg@16 Accuracy
2
AIME 2024
60.9Avg@16 Accuracy
2
AIME 24
3.41Match Rate (MAT)
2
AIME 25
30Total Count
2
AIME 24
30Total Score
2
AIME 25
21.6Accuracy
2
AIME 2024
61.67Avg@32 Accuracy
2
AIME 24
3.82Mean Accepted Tokens
2
AIME 25
40.8Accuracy
2
AIME24
57.1Accuracy
2
AIME MEDIUM reasoning with tools 2025
91.7Score
2
AIME 24-25
26.4Score
2