Benchmarks / Math

FrontierMath Tiers 1–3 (v2)

Epoch AI's FrontierMath Tiers 1 3, version 2: expert-written mathematics problems evaluated with Python tools. Accuracy on the private set, not the public sample questions. Data: Epoch AI, CC BY.

84models
10%Open Index weight
–last updated
Epoch AIpublisher

FrontierMath Tiers 1–3 (v2)

Higher is better · best effort per model

Chart settings
84 of 84 selected

full table below

#ModelEffortScoreCost / taskHarness
1GPT-6.1 Solmax93.7%–Epoch AI
2GPT-6 Astramax93.7%–Epoch AI
3Opus 5.5max91.2%–Epoch AI
4Fable 5.1max90.2%–Epoch AI
5GPT-6 Solmax89.8%–Epoch AI
6GPT-5.6 Solmax89.1%–Epoch AI
7Sonnet 5.5max88.8%–Epoch AI
8GPT-5.5 Proxhigh87.7%–Epoch AI
9Fable 5max87.0%–Epoch AI
10Gemini 4high86.0%–Epoch AI
11GPT-5.6 Terramax86.0%–Epoch AI
12Opus 5max85.6%–Epoch AI
13GPT-5.5xhigh85.3%–Epoch AI
14GPT-5.4 Proxhigh82.5%–Epoch AI
15GPT-5.6 Lunamax82.1%–Epoch AI
16Opus 4.8max80.0%–Epoch AI
17GPT-6 Lunamax78.9%–Epoch AI
18GPT-5.4xhigh78.6%–Epoch AI
19Haiku 5.5max75.1%–Epoch AI
20Muse Spark 1.3xhigh74.4%–Epoch AI
21GPT-5.2 Proxhigh74.0%–Epoch AI
22Kimi K3max72.2%–Epoch AI
23Gemini 3.7 Flashhigh71.6%–Epoch AI
24Opus 4.7max70.2%–Epoch AI
25GLM 5.3max68.8%–Epoch AI
26Gemini 3.8 Flashhigh68.4%–Epoch AI
27GPT-5.2xhigh67.4%–Epoch AI
28DeepSeek V4.1 Flashmax67.4%–Epoch AI
29Grok 4.6xhigh66.0%–Epoch AI
30Opus 4.6max66.0%–Epoch AI
31Qwen 3.8 Maxxhigh65.6%–Epoch AI
32Sonnet 5max65.6%–Epoch AI
33DeepSeek V4 Pro 0813max64.6%–Epoch AI
34Qwen 3.7 Maxmax64.6%–Epoch AI
35Gemini 3.5 Flashhigh62.8%–Epoch AI
36Gemini 3.1 Prohigh59.6%–Epoch AI
37GLM 5.2max59.2%–Epoch AI
38Gemini 3.6 Flashhigh58.9%–Epoch AI
39Mistral Large 4high58.6%–Epoch AI
40DeepSeek V4 Flash 0731max57.5%–Epoch AI
41Grok 4.5high57.2%–Epoch AI
42Kimi K2.6max57.2%–Epoch AI
43GLM 5.3 Flashmax55.8%–Epoch AI
44GPT-5 Prohigh55.8%–Epoch AI
45GPT-5high55.4%–Epoch AI
46Kimi K2.7 Codemax54.0%–Epoch AI
47Grok 4.7xhigh53.0%–Epoch AI
48GPT-5.4 Minixhigh51.2%–Epoch AI
49Gemini 3 Flashhigh51.2%–Epoch AI
50GPT-5 Minihigh46.7%–Epoch AI
51Inkling Smallxhigh46.3%–Epoch AI
52DeepSeek V4 Promax45.3%–Epoch AI
53GPT-5.4 Nanohigh44.9%–Epoch AI
54Grok 4.20xhigh44.9%–Epoch AI
55Grok 4.3high42.8%–Epoch AI
56Qwen 3.6 Plusmax38.2%–Epoch AI
57GLM 5.1max36.8%–Epoch AI
58o4 minihigh36.1%–Epoch AI
59Qwen 3.6 27Bmax35.1%–Epoch AI
60Qwen 3.7 Plusnone34.4%–Epoch AI
61Opus 4.5high34.4%–Epoch AI
62o3high33.3%–Epoch AI
63Inklingxhigh33.3%–Epoch AI
64Qwen 3.5 397B-A17Bnone31.2%–Epoch AI
65Gemini 3.1 Flash-Litehigh27.7%–Epoch AI
66GPT-5.5 Instant–26.3%–Epoch AI
67Gemini 3.5 Flash-Litehigh26.0%–Epoch AI
68Gemini 2.5 Prohigh24.6%–Epoch AI
69Sonnet 4.5high23.9%–Epoch AI
70Qwen 3.6 Flashnone22.5%–Epoch AI
71Qwen 3.6 35B-A3Bnone20.4%–Epoch AI
72GPT-5 Nanohigh20.0%–Epoch AI
73Qwen 3.7 Flashnone19.3%–Epoch AI
74Qwen 3 Maxmax18.9%–Epoch AI
75o3 minihigh18.6%–Epoch AI
76Qwen 3.5 Flashnone18.2%–Epoch AI
77o1high14.7%–Epoch AI
78Opus 4.1high12.6%–Epoch AI
79GPT-4.1 Minihigh6.7%–Epoch AI
80GPT-4.1high6.0%–Epoch AI
81GPT-4o minihigh0.7%–Epoch AI
82GPT-4high0.7%–Epoch AI
83GPT-4ohigh0.4%–Epoch AI
84GPT-3.5high0.0%–Epoch AI