AIME
Canonical page on the main site: sinoaihub.com/benchmarks/aime
Type
benchmark
Key facts
- Task type: Advanced competition mathematics — 15 integer-answer problems
- Dataset size: 15 problems per exam (integer answers from 000 to 999); benchmark scores typically report AIME 2024 or AIME 2025
- Evaluation method: Models answer each problem; graded by exact match against the integer answer
- Scoring: Accuracy (% of 15 problems answered correctly); the raw exam score is 0–15
- Recorded evaluations: 0
Description
The American Invitational Mathematics Examination, repurposed as a benchmark: 15 integer-answer math problems used to test a model's advanced mathematical reasoning, with scores typically reported for AIME 2024 or AIME 2025.
Evaluations
No evaluations recorded — no tracked Chinese model publishes an AIME score as of 2026-09-29.
Methodology
Task type: Advanced competition mathematics — 15 integer-answer problems
Dataset size: 15 problems per exam (integer answers from 000 to 999); benchmark scores typically report AIME 2024 or AIME 2025
Evaluation method: Models answer each problem; graded by exact match against the integer answer
Scoring: Accuracy (% of 15 problems answered correctly); the raw exam score is 0–15
Contamination notes: AIME is a fixed public exam, so models trained after a given year may have seen its problems; vendors differ in which year(s) they report.
Relevant Models
No tracked model currently publishes an AIME score. Natural candidates in the database include DeepSeek-V4-Pro, Kimi K3 and Qwen3.8-Max.
Limitations
AIME is an exam, not a versioned benchmark, so scores are only comparable when the same year is reported. Exact-match integer grading gives no partial credit for correct reasoning with a wrong final answer. No Chinese model in the China AI Hub database currently publishes an AIME score, so this page records no evaluations.
Verification Status
verified
Benchmark Changes
No documented benchmark changes on record as of 2026-09-29.
Last Verified
2026-09-29
Source history
No documented source-change events located as of 2026-09-29.
Sources
| evidence_id | source_name | source_url | source_type | published | verified | confidence | conflict |
|---|---|---|---|---|---|---|---|
| src-benchmarks-aime-1 | Mathematical Association of America — American Invitational Mathematics Examination (AIME) | https://maa.org/math-competitions/american-invitational-mathematics-examination-aime | Official documentation | — | 2026-09-29 | high | — |
| src-benchmarks-aime-2 | Mathematical Association of America — Competitions | https://maa.org/math-competitions | Official documentation | — | 2026-09-29 | high | — |