Skip to content

Benchmark

mu-MATH accuracy

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

Models measured
17
Scores recorded
85
Sources
1
IQ weight
1.63

Score distribution

75%
Mean
76%
Median
9.8%
Std dev
61.2%
10th percentile
87%
90th percentile
48.7%
Min
48.7%94.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.63 for chat models, measured over 17 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1o1Llama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH94.2%
2o1Gemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH91.3%
3O1 MiniLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH91.3%
4DeepSeek R1Llama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH89.9%
5o1mu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH89.5%
6o1GPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH88.4%
7Gemini 2.0 Flash ExperimentalGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH87%
8O1 MiniGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87%
9QwQ 32B PreviewQwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87%
10Gemini 2.0 Flash ExperimentalLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH85.5%
11QwQ 32B PreviewLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH85.5%
12Gemini 1.5 ProLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH85.2%
13O1 Minimu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.8%
14o1Qwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.1%
15GPT 4oLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.8%
16QwQ 32B Previewmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.3%
17Gemini 1.5 FlashLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83%
18DeepSeek R1Gemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.6%
19QwQ 32B PreviewGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.6%
20DeepSeek R1mu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.2%
21Gemini 1.5 ProGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH81.5%
22Claude 3.5 SonnetLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH81.2%
23Gemini 2.0 Flash Experimentalmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH81.2%
24O1 MiniGPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH81.2%
25Qwen2.5 72B InstructLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.8%
26Gemini 1.5 Promu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.7%
27Mistral Large Instruct 2411Llama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.1%
28DeepSeek R1GPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH79.7%
29O1 MiniQwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH79.7%
30Qwen2.5 Math 72B InstructGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH79.7%
31GPT-4o miniLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH79.3%
32QwQ 32B PreviewGPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH78.3%
33Gemini 1.5 ProGPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH78.2%
34Qwen2.5 Math 72B InstructLlama-3.1-70B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH78.2%
35Gemini 1.5 ProQwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.9%
36Mistral Large Instruct 2411Gemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.9%
37GPT 4oGPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.5%
38GPT 4omu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.4%
39Qwen2.5 72B InstructGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.1%
40DeepSeek R1Qwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH76.8%
41Gemini 2.0 Flash ExperimentalQwen2.5-72B-Instruct, accuracy, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH76.8%
42Mistral Large Instruct 2411mu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH76.7%
43Mistral Large Instruct 2411GPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH76%
44Qwen2.5 72B Instructmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH75.7%
45Gemini 2.0 Flash ExperimentalGPT-4o, accuracy, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH75.4%
46Claude 3.5 SonnetGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH75.3%
47Claude 3.5 Sonnetmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH75%
48Gemini 1.5 FlashGemini-1.5-Pro, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH74.9%
49Gemini 1.5 Flashmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH74.9%
50Qwen2.5 Math 72B Instructmu_math, accuracy, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH74.4%

Showing 1–50 of 85 results.

Data credits: Toloka, U-MATH and mu-MATH public leaderboard. MIT Space; generated solution outputs retain underlying model terms and are not ingested. (licence)