Skip to content

Benchmark

FrontierMath Tier 4

Math benchmarks measure how well models solve maths problems. Scores are percentages.

Models measured
47
Scores recorded
62
Sources
1
IQ weight
3.01

Score distribution

7.5%
Mean
4.2%
Median
7.8%
Std dev
0%
10th percentile
18.8%
90th percentile
0%
Min
0%31.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 3.01 for chat models, measured over 47 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 4.8max· Epoch AI Benchmarking Hub data31.3%
2GPT 5.4xhigh· Epoch AI Benchmarking Hub data27.1%
3Claude Opus 4.7xhigh· Epoch AI Benchmarking Hub data22.9%
4Claude Opus 4.6max· Epoch AI Benchmarking Hub data22.9%
5Claude Opus 4.632k· Epoch AI Benchmarking Hub data20.8%
6Claude Opus 4.664k· Epoch AI Benchmarking Hub data20.8%
7GPT 5.2high· Epoch AI Benchmarking Hub data18.8%
8GPT 5.2xhigh· Epoch AI Benchmarking Hub data18.8%
9Gemini 3 Pro Preview· Epoch AI Benchmarking Hub data18.8%
10Gemini 3.1 Pro Preview· Epoch AI Benchmarking Hub data16.7%
11GPT 5.2medium· Epoch AI Benchmarking Hub data16.7%
12GPT-5 Prohigh· Epoch AI Benchmarking Hub data14.6%
13Muse Spark· Epoch AI Benchmarking Hub data14.6%
14Claude Opus 4.6· Epoch AI Benchmarking Hub data14.6%
15Gemini 3.5 Flashhigh· Epoch AI Benchmarking Hub data14.6%
16Kimi K2.6· Epoch AI Benchmarking Hub data14.6%
17GLM 5.1· Epoch AI Benchmarking Hub data12.5%
18GPT 5high· Epoch AI Benchmarking Hub data12.5%
19GPT 5.1high· Epoch AI Benchmarking Hub data12.5%
20Qwen3.6 Plus· Epoch AI Benchmarking Hub data8.3%
21Claude Sonnet 4.616k· Epoch AI Benchmarking Hub data8.3%
22GPT 5medium· Epoch AI Benchmarking Hub data6.3%
23GPT 5.2low· Epoch AI Benchmarking Hub data6.3%
24GPT 5.4 Nanohigh· Epoch AI Benchmarking Hub data6.3%
25GPT-5 minihigh· Epoch AI Benchmarking Hub data6.3%
26o4 Minihigh· Epoch AI Benchmarking Hub data6.3%
27Kimi K2.5· Epoch AI Benchmarking Hub data4.2%
28Claude Opus 427k· Epoch AI Benchmarking Hub data4.2%
29Claude Opus 4.127k· Epoch AI Benchmarking Hub data4.2%
30Claude Opus 4.5· Epoch AI Benchmarking Hub data4.2%
31Claude Opus 4.532k· Epoch AI Benchmarking Hub data4.2%
32Claude Sonnet 4.532k· Epoch AI Benchmarking Hub data4.2%
33Gemini 2.5 Flash· Epoch AI Benchmarking Hub data4.2%
34Gemini 2.5 Pro· Epoch AI Benchmarking Hub data4.2%
35Gemini 3 Flash· Epoch AI Benchmarking Hub data4.2%
36GPT 5.1medium· Epoch AI Benchmarking Hub data4.2%
37GPT-5 minimedium· Epoch AI Benchmarking Hub data4.2%
38o3 Minihigh· Epoch AI Benchmarking Hub data4.2%
39Qwen3.6 Max Preview· Epoch AI Benchmarking Hub data4.2%
40GLM 4.6· Epoch AI Benchmarking Hub data2.1%
41DeepSeek V3.2· Epoch AI Benchmarking Hub data2.1%
42GLM 5· Epoch AI Benchmarking Hub data2.1%
43Claude Haiku 4.532k· Epoch AI Benchmarking Hub data2.1%
44Claude Opus 4.516k· Epoch AI Benchmarking Hub data2.1%
45Claude Sonnet 4.5· Epoch AI Benchmarking Hub data2.1%
46GPT-5 nanomedium· Epoch AI Benchmarking Hub data2.1%
47Grok 4· Epoch AI Benchmarking Hub data2.1%
48o4 Minimedium· Epoch AI Benchmarking Hub data2.1%
49OpenAI o3high· Epoch AI Benchmarking Hub data2.1%
50Qwen3.5 Plus· Epoch AI Benchmarking Hub data2.1%

Showing 1–50 of 62 results.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence)