← OpenCompass leaderboards

AIME 2025

OpenCompass · 48 models

RankModelScore
1DeepSeek-V3.2-Speciale96
2GLM-595.8
3Step-3.5-Flash95.7
4GLM-4.795.4
5GPT-5-2025-08-07 (high)94.1
5Kimi-K2-Thinking94.1
7Gemini-3-Pro-Preview93.4
7gpt-oss-120b (high)93.4
9DeepSeek-V3.293
10MiMo-V2-Flash92.9
11Qwen3.5-397B-A17B92.3
12Kimi-K2.591.9
13o4-mini-2025-04-16 (high)91
14GPT-5-mini-2025-08-07 (high)90.9
14Qwen3-235B-A22B-Thinking-250790.9
16GLM-4.690.3
17DeepSeek-R1-052889
17Qwen3-Next-80B-A3B-Thinking89
19Gemini-2.5-Pro88.7
20gpt-oss-20b (high)87.9
21Qwen3-30B-A3B-Thinking-250786.8
22MiniMax-M2.586.2
23Seed-OSS-36B-Instruct86
24GLM-4.585.8
25GPT-5-nano-2025-08-07 (high)85.4
26o3-2025-04-1685.1
27doubao-seed-1.6-thinking-25061582.8
28MiniCPM-SALA81
29Qwen3-4B-Thinking-250780
30MiniMax-M279.1
31TeleChat-thinking79
32ERNIE-4.5-21B-A3B-Thinking76.2
33Qwen3-32B (Thinking)70.3
34Finix-P1-32B (Thinking)70.2
35Qwen3-235B-A22B-Instruct-250769.5
36Qwen3-Next-80B-A3B-Instruct69.2
37Claude Sonnet 4 (Thinking)68.7
38Qwen3-8B (Thinking)66.2
39JT-Think65.9
40Hunyuan-A13B-Instruct65.7
41Qwen3-30B-A3B-Instruct-250763.8
42LongCat-Flash-Chat61
43Kimi-K2-Instruct49.1
44Qwen3-4B-Instruct-250746.9
45GPT-4.1-2025041435.3
46ERNIE-4.5-300B-A47B-PT32.6
47Gemma-3-27B-it22.4
48Llama4-Maverick-17B-128E-Instruct15.2