Benchmarks / Math

AIME (Vals)

Challenging national math exam given to top high-school students Scores are the values published by VALS AI. Source benchmark version 1.

80models
–Open Index weight
2026-04-16last updated
Vals AIpublisher

AIME (Vals)

Higher is better · best effort per model

Chart settings
80 of 80 selected

full table below

#ModelEffortScoreCost / taskHarness
1Gemini 3.1 Prohigh98.1%$0.127–
2GPT-5.2xhigh96.9%$0.135–
3Muse Spark–96.9%$0.0046–
4Gemini 3 Prohigh96.7%$0.169–
5GPT-5.4xhigh96.7%$0.138–
6Grok 4.20xhigh96.5%$0.059–
7Opus 4.7max96.3%$0.451–
8Gemini 3 Flashhigh95.6%$0.046–
9Opus 4.6max95.6%$0.344–
10GPT-5.4 Minixhigh95.6%$0.129–
11Kimi K2.5max95.6%$0.064–
12Opus 4.5high95.4%$1.32–
13Qwen 3.6 Plusmax94.6%$0.094–
14GPT-5high93.4%$0.093–
15GPT-5.1high93.3%$0.150–
16GLM 4.7none93.3%$0.043–
17GLM 4.6none92.7%$0.116–
18gpt-oss-120bhigh92.6%$0.010–
19Qwen 3.5 Flashmax92.5%$0.011–
20Sonnet 4.6max92.3%$0.436–
21Grok 4.1 Fasthigh91.9%$0.0041–
22GLM 5.1max91.9%$0.071–
23GLM 5high91.7%$0.070–
24GPT-5 Minihigh91.5%$0.023–
25Grok 4 Fastnone91.3%$0.0043–
26MiniMax M2.7high91.0%$0.023–
27Grok 4high90.6%$0.174–
28GPT-5.4 Nanohigh88.8%$0.0034–
29MiniMax M2.5high88.8%$0.041–
30Sonnet 4.5high88.2%$0.267–
31GLM 4.5xhigh86.7%$0.047–
32o3 minihigh86.5%$0.051–
33Qwen 3.5 Plusnone86.0%$0.045–
34gpt-oss-20bhigh86.0%$0.011–
35Gemini 2.5 Prohigh85.8%$0.027–
36Kimi K2medium85.4%$0.045–
37o3high85.3%$0.060–
38Grok 3 Minihigh85.0%$0.0083–
39DeepSeek V3.2high84.6%$0.0044–
40Qwen 3 235B–84.0%$0.012–
41o4 minihigh83.7%$0.026–
42Magistral Medium 1.2medium83.5%$0.087–
43Gemini 3.1 Flash-Litehigh83.3%$0.0070–
44Haiku 4.5xhigh82.7%$0.174–
45GPT-5 Nanohigh81.2%$0.0089–
46Qwen 3 Maxmax81.0%$0.033–
47Magistral Small 1.2–80.7%$0.025–
48Opus 4.1high78.2%$0.860–
49MiniMax M2.1–77.9%$0.024–
50Sonnet 4high76.3%$0.295–
51DeepSeek R1xhigh74.0%$0.053–
52o1high71.5%$0.659–
53Grok 3none58.8%$0.091–
54Llama 3.3none53.5%$0.0000–
55DeepSeek V3none52.2%$0.0032–
56Gemini 2.5 Flashmedium51.5%$0.0050–
57GPT-4.1 Minihigh49.4%$0.0079–
58Sonnet 3.7medium44.6%$0.395–
59Mistral Large 3none42.9%$0.0061–
60Mistral Medium 3.1none42.3%$0.0000–
61Gemini 2.5 Flash Litenone42.1%$0.0006–
62Opus 4medium41.3%$0.110–
63GPT-4.1high39.6%$0.050–
64Gemini 2.0 Flashmedium29.8%$0.0010–
65GPT-4.1 Nanohigh26.5%$0.0012–
66Llama 4 Maverickmedium25.2%$0.0011–
67Llama 4 Scoutmedium19.0%$0.0008–
68Gemini 1.5 Pronone18.8%$0.0042–
69Gemini 1.5 Flashnone17.3%$0.0003–
70Grok 2none15.2%$0.015–
71GPT-4ohigh14.0%$0.012–
72Command Anone13.3%$0.012–
73GPT-4o minihigh11.5%$0.0010–
74Claude 3.5 Sonnet Latest–10.0%$0.0088–
75Mistral Large 11–9.2%$0.0063–
76Mistral Small 02–5.6%$0.0008–
77Mistral Small 3.1none3.5%$0.0003–
78Haiku 3.5medium3.3%$0.0024–
79Jamba 1.6 Mininone0.4%$0.0005–
80Jamba 1.6 Largenone0.4%$0.0077–