Skip to content

Benchmark

MATH Level 5

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

Models measured
221
Scores recorded
238
Sources
2
IQ weight
1.72

Score distribution

32.6%
Mean
22%
Median
30.4%
Std dev
2.3%
10th percentile
87.1%
90th percentile
0%
Min
0%98.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.72 for chat models, measured over 221 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1GPT 5high· Epoch AI Benchmarking Hub data98.1%
2GPT 5medium· Epoch AI Benchmarking Hub data97.9%
3GPT-5 minihigh· Epoch AI Benchmarking Hub data97.8%
4o4 Minihigh· Epoch AI Benchmarking Hub data97.8%
5OpenAI o3high· Epoch AI Benchmarking Hub data97.8%
6Claude Sonnet 4.532k· Epoch AI Benchmarking Hub data97.7%
7Qwen3 Max· Epoch AI Benchmarking Hub data97.1%
8GPT-5 minimedium· Epoch AI Benchmarking Hub data96.8%
9DeepSeek R1· Epoch AI Benchmarking Hub data96.6%
10o3 Minihigh· Epoch AI Benchmarking Hub data96.5%
11Claude Haiku 4.532k· Epoch AI Benchmarking Hub data96.4%
12Gemini 2.5 Pro Preview 05-06· Epoch AI Benchmarking Hub data95.9%
13Gemini 2.5 Pro Preview 03-25· Epoch AI Benchmarking Hub data95.6%
14GPT-5 nanomedium· Epoch AI Benchmarking Hub data95.2%
15o3 Minimedium· Epoch AI Benchmarking Hub data95.2%
16GPT-5 nanohigh· Epoch AI Benchmarking Hub data94.9%
17o1high· Epoch AI Benchmarking Hub data94.7%
18o1medium· Epoch AI Benchmarking Hub data94.4%
19Claude Sonnet 3.764k· Epoch AI Benchmarking Hub data91.2%
20Grok 3 Minilow, beta· Epoch AI Benchmarking Hub data90.9%
21Claude Sonnet 3.732k· Epoch AI Benchmarking Hub data90%
22DeepSeek R1 Distill Llama 70b· Epoch AI Benchmarking Hub data89.9%
23O1 Minihigh· Epoch AI Benchmarking Hub data89.2%
24Grok 3beta· Epoch AI Benchmarking Hub data88.7%
25Grok 3 Minihigh, beta· Epoch AI Benchmarking Hub data88.1%
26GPT-4.1 mini· Epoch AI Benchmarking Hub data87.3%
27DeepSeek R1 Distill Qwen 14B· Epoch AI Benchmarking Hub data
Values from 2 sources
87.1%
28Claude Haiku 4.5· Epoch AI Benchmarking Hub data86.9%
29Claude Sonnet 3.716k· Epoch AI Benchmarking Hub data86.3%
30Claude Opus 4· Epoch AI Benchmarking Hub data85%
31Claude Sonnet 4· Epoch AI Benchmarking Hub data84.4%
32O1 Minimedium· Epoch AI Benchmarking Hub data84.3%
33gemini-2.0-pro-exp-02-05· Epoch AI Benchmarking Hub data83.5%
34GPT 4.1· Epoch AI Benchmarking Hub data83%
35Gemini 2.0 Flash· Epoch AI Benchmarking Hub data82.2%
36O1 Preview· Epoch AI Benchmarking Hub data81.6%
37Mistral Medium 3· Epoch AI Benchmarking Hub data81.6%
38GPT-4.5 Preview· Epoch AI Benchmarking Hub data78.6%
39DeepSeek V3· Epoch AI Benchmarking Hub data75.5%
40Gemma 3 27B IT· Epoch AI Benchmarking Hub data74%
41Llama 4 Maverick· Epoch AI Benchmarking Hub data73%
42gemini-1.5-pro-002· Epoch AI Benchmarking Hub data70.4%
43GPT-4.1 nano· Epoch AI Benchmarking Hub data70%
44Qwen 3 235B A22B· Epoch AI Benchmarking Hub data68.9%
45Claude Sonnet 3.7· Epoch AI Benchmarking Hub data68.2%
46Qwen Max· Epoch AI Benchmarking Hub data67.2%
47Qwen Plus· Epoch AI Benchmarking Hub data65.3%
48Phi 4· Epoch AI Benchmarking Hub data
Values from 2 sources
64.9%
49Grok 2· Epoch AI Benchmarking Hub data63.5%
50Qwen2.5 72B Instruct· Epoch AI Benchmarking Hub data
Values from 2 sources
63.2%

Showing 1–50 of 238 results.

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