Math

IMOAnswerBench leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on IMOAnswerBench is 90.9% by dots3-note Preview. 15 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	dots3-note Preview	90.9%	measured
2	Qwen3.7 Max	90.0%	measured
3	DeepSeek V4 Pro 0813	89.8%	measured
4	DeepSeek V4 Flash 0731	88.4%	measured
5	Qwen3.7 Plus	86.0%	measured
6	Solar Open 2	85.7%	estimated ± 5.9 pp, medium confidence
7	Kimi K2.6	85.1%	estimated ± 5.9 pp, medium confidence
8	GLM-5.2	85.0%	estimated ± 5.9 pp, medium confidence
9	Inkling-Small	83.7%	estimated ± 5.9 pp, medium confidence
10	Ling 3.0 Flash	83.7%	measured
11	Qwen3.5 397B	82.5%	estimated ± 5.9 pp, medium confidence
12	Qwen3.6 Plus	82.4%	estimated ± 5.9 pp, medium confidence
13	Kimi K2.5	82.0%	estimated ± 5.9 pp, medium confidence
14	GLM-5	81.7%	estimated ± 5.9 pp, medium confidence
15	Claude Opus 4.5	81.0%	estimated ± 5.9 pp, medium confidence
16	MAI-Thinking-1	80.8%	estimated ± 5.9 pp, medium confidence
17	Qwen3.6-27B	80.5%	estimated ± 5.9 pp, medium confidence
18	Qwen3.6-35B-A3B	80.1%	estimated ± 5.9 pp, medium confidence
19	GLM-5.1	79.5%	estimated ± 5.9 pp, medium confidence
20	K-EXAONE 2.0	78.6%	measured
21	MiniCPM5-2B	67.5%	estimated ± 5.9 pp, medium confidence
22	ZAYA1-8B	59.3%	measured
23	LongCat-Flash-Lite-Sparse	49.4%	measured
24	MiniCPM5-1B	34.1%	estimated ± 5.9 pp, low confidence

IMOAnswerBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General