Skip to content

Benchmark

FrontierMath

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

Models measured
71
Scores recorded
109
Sources
2
IQ weight
1.73

Score distribution

19.5%
Mean
16.4%
Median
18%
Std dev
1.4%
10th percentile
40.3%
90th percentile
0%
Min
0%89%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.73 for chat models, measured over 71 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51o4 Minimedium· Epoch AI Benchmarking Hub data19%
52OpenAI o3high· Epoch AI Benchmarking Hub data18.7%
53GPT 5.1low· Epoch AI Benchmarking Hub data17.3%
54OpenAI o3medium· Epoch AI Benchmarking Hub data16.9%
55GLM 5· Epoch AI Benchmarking Hub data16.4%
56OpenAI o3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim15.8%
57Claude Sonnet 4.532k· Epoch AI Benchmarking Hub data15.2%
58Gemini 2.5 Pro· Epoch AI Benchmarking Hub data14.1%
59Claude Sonnet 4.559k· Epoch AI Benchmarking Hub data13.5%
60o3 Minihigh· Epoch AI Benchmarking Hub data12.4%
61o4 Minilow· Epoch AI Benchmarking Hub data10.7%
62Qwen3.6 Flash· Epoch AI Benchmarking Hub data10.3%
63OpenAI o3low· Epoch AI Benchmarking Hub data9.7%
64GPT-5 nano· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim9.6%
65Claude Sonnet 4.5· Epoch AI Benchmarking Hub data9.3%
66o1high· Epoch AI Benchmarking Hub data9.3%
67o3 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim9.2%
68Qwen 3 235B A22B Thinking 2507· Epoch AI Benchmarking Hub data8.5%
69GPT-5 nanohigh· Epoch AI Benchmarking Hub data8.3%
70o3 Minimedium· Epoch AI Benchmarking Hub data8.1%
71Claude Opus 4.127k· Epoch AI Benchmarking Hub data7.2%
72GPT-5 nanomedium· Epoch AI Benchmarking Hub data7.2%
73Qwen3.5 Flash· Epoch AI Benchmarking Hub data6.2%
74Claude Haiku 4.532k· Epoch AI Benchmarking Hub data5.9%
75Claude Opus 4.1· Epoch AI Benchmarking Hub data5.9%
76Grok 3 Minihigh, beta· Epoch AI Benchmarking Hub data5.9%
77GPT 4.1· Epoch AI Benchmarking Hub data5.5%
78o1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim5.5%
79Gemini 2.5 Flash· Epoch AI Benchmarking Hub data4.8%
80Claude Opus 4· Epoch AI Benchmarking Hub data4.5%
81GPT-4.1 mini· Epoch AI Benchmarking Hub data4.5%
82Claude Haiku 4.5· Epoch AI Benchmarking Hub data4.1%
83Claude Opus 427k· Epoch AI Benchmarking Hub data4.1%
84Claude Sonnet 3.716k· Epoch AI Benchmarking Hub data4.1%
85Claude Sonnet 4· Epoch AI Benchmarking Hub data4.1%
86GLM 4.6· Epoch AI Benchmarking Hub data3.8%
87Grok 3beta· Epoch AI Benchmarking Hub data3.8%
88Claude Sonnet 3.732k· Epoch AI Benchmarking Hub data3.4%
89Claude Sonnet 3.7· Epoch AI Benchmarking Hub data3.1%
90Claude Sonnet 3.764k· Epoch AI Benchmarking Hub data3.1%
91Grok 3 Minilow, beta· Epoch AI Benchmarking Hub data2.8%
92GLM 4.7· Epoch AI Benchmarking Hub data2.4%
93GPT 5.1none· Epoch AI Benchmarking Hub data2.1%
94DeepSeek V3· Epoch AI Benchmarking Hub data1.7%
95Gemini 2.0 Flash· Epoch AI Benchmarking Hub data1.7%
96O1 Minimedium· Epoch AI Benchmarking Hub data1.7%
97Qwen Plus· Epoch AI Benchmarking Hub data1.7%
98O1 Minihigh· Epoch AI Benchmarking Hub data1.4%
99Claude 3.5 Sonnet· Epoch AI Benchmarking Hub data1%
100GPT-4.1 nano· Epoch AI Benchmarking Hub data1%

Showing 51–100 of 109 results.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats