benchgap
Math

IMOAnswerBench leaderboard

As of 2026-11-21, the highest measured score on IMOAnswerBench is 90.9% by dots3-note Preview. 29 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1dots3-note Preview90.9%measured
2Qwen3.7 Max90.0%measured
3DeepSeek V4 Pro89.8%measured
4DeepSeek V4 Flash 073188.4%measured
5Inkling 2 Small86.3%estimated ± 5.6 pp, medium confidence
6Qwen3.7 Plus86.0%measured
7Solar Open 285.3%estimated ± 5.6 pp, medium confidence
8Kimi K2.684.7%estimated ± 5.6 pp, medium confidence
9GLM-5.284.6%estimated ± 5.6 pp, medium confidence
10Ling 3.0 Flash83.7%measured
11Ripple-65B Preview83.5%estimated ± 5.6 pp, medium confidence
12Inkling-Small83.3%estimated ± 5.6 pp, medium confidence
13Qwen3.5 397B82.1%estimated ± 5.6 pp, medium confidence
14Qwen3.6 Plus82.0%estimated ± 5.6 pp, medium confidence
15Kimi K2.581.6%estimated ± 5.6 pp, medium confidence
16GLM-581.2%estimated ± 5.6 pp, medium confidence
17Claude Opus 4.580.6%estimated ± 5.6 pp, medium confidence
18MAI-Thinking-180.4%estimated ± 5.6 pp, medium confidence
19Qwen3.6-27B80.1%estimated ± 5.6 pp, medium confidence
20Qwen3.6-35B-A3B79.7%estimated ± 5.6 pp, medium confidence
21GLM-5.179.1%estimated ± 5.6 pp, medium confidence
22K-EXAONE 2.078.6%measured
23Beam 1.0 501B-A23B FP875.6%estimated ± 10.7 pp, low confidence
24Beam 1.0 501B-A23B75.4%estimated ± 10.7 pp, low confidence
25Beam 1.0 501B-A23B NVFP475.3%estimated ± 10.7 pp, low confidence
26Spark-X2.6-4B75.2%measured
27A.X K275.1%estimated ± 10.7 pp, low confidence
28Inkling75.1%estimated ± 10.7 pp, low confidence
29Ternary Bonsai 2 27B74.5%estimated ± 10.7 pp, low confidence
30Solar Pro 474.3%estimated ± 10.7 pp, low confidence
31Muse Glimmer 30B74.0%estimated ± 10.7 pp, low confidence
32MiniMax M2.773.1%estimated ± 5.6 pp, medium confidence
33MiniCPM5-2B67.1%estimated ± 5.6 pp, medium confidence
34Gemma 4 12B64.9%estimated ± 10.7 pp, low confidence
35ZAYA1-74B-Preview64.3%estimated ± 10.7 pp, low confidence
36ZAYA1-8B59.3%measured
37LongCat-Flash-Lite-Sparse49.4%measured
38LFM2.5-8B-A1B47.3%estimated ± 10.7 pp, low confidence
39MiniCPM5-1B33.9%estimated ± 5.6 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General