Skip to content

Benchmark

mu-MATH TNR

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
2.67

Score distribution

75.3%
Mean
80.6%
Median
13.9%
Std dev
54%
10th percentile
90.2%
90th percentile
41.2%
Min
41.2%95.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.67 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, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH95.8%
2O1 MiniLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH93.8%
3Claude 3.5 SonnetLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH93.1%
4Gemini 1.5 FlashLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH92%
5DeepSeek R1Llama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH91.7%
6Gemini 1.5 FlashGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH91.5%
7Gemini 1.5 ProLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH90.8%
8GPT 4oLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH90.8%
9GPT-4o miniLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH90.2%
10Claude 3.5 Sonnetmu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH89.5%
11Claude 3.5 SonnetGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH88.5%
12o1mu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH88.4%
13Gemini 1.5 Flashmu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH88.3%
14GPT-4o minimu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH88.1%
15Claude 3.5 SonnetQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.9%
16DeepSeek R1GPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.9%
17o1GPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.9%
18GPT-4o miniGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.8%
19DeepSeek R1mu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.7%
20Gemini 2.0 Flash ExperimentalLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH87.5%
21GPT-4o miniQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH87.1%
22GPT-4o miniGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH86.3%
23O1 Minimu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH86.2%
24GPT 4omu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH85.9%
25Claude 3.5 SonnetGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH85.4%
26Gemini 1.5 FlashQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH85.3%
27O1 MiniGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.8%
28DeepSeek R1Gemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.6%
29o1Gemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.6%
30O1 MiniGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.6%
31Gemini 1.5 Promu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84.5%
32Gemini 1.5 FlashGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH84%
33DeepSeek R1Qwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.9%
34Qwen2.5 72B InstructLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.9%
35GPT 4oQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.6%
36QwQ 32B PreviewLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.3%
37GPT 4oGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH83.2%
38Gemini 1.5 ProGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.9%
39GPT 4oGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.9%
40Mistral Large Instruct 2411Llama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH82.8%
41Gemini 1.5 ProQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH81%
42o1Qwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.6%
43QwQ 32B PreviewQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.6%
44Gemini 1.5 ProGPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH80.2%
45Mistral Large Instruct 2411mu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.7%
46O1 MiniQwen2.5-72B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH77.4%
47Gemini 2.0 Flash ExperimentalGemini-1.5-Pro, TNR, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH76.9%
48Qwen2.5 Math 72B InstructLlama-3.1-70B-Instruct, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH76.4%
49Mistral Large Instruct 2411GPT-4o, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH76.3%
50QwQ 32B Previewmu_math, TNR, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH75.4%

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)