Skip to content

Benchmark

U-MATH sequences_series accuracy

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

Models measured
30
Scores recorded
90
Sources
1
IQ weight
1.63

Score distribution

44.8%
Mean
48.7%
Median
27.3%
Std dev
0%
10th percentile
84.4%
90th percentile
0%
Min
0%94%
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 30 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Nvidia Llama 3.1 Nemotron 70B Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH42.2%
52GPT-4o minisequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH40.3%
53GPT-4o minisequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH40%
54Claude 3.5 Sonnetsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH38.7%
55Qwen2.5 7B Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH38.7%
56Claude 3.5 Sonnetsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH37.7%
57Qwen2.5 7B Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH37.7%
58Llama 3.2 90B Vision Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH36%
59Llama 3.2 90B Vision Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH35.7%
60Llama 3.1 70B Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH34.4%
61Llama 3.1 70B Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH34%
62Qwen2-VL (72B) Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH34%
63Qwen2-VL (72B) Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH33.8%
64Llama 3.1 8B Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH28%
65Llama 3.1 8B Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH27.3%
66Gemini 2.0 Flash Experimentalsequences_series, visual, gpt-4o-2024-08-06, thinking· U-MATH and mu-MATH25%
67Llama 3.2 90B Vision Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH25%
68Llama 3.3 70B Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH25%
69Mistral Large Instruct 2411sequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH25%
70Qwen2-VL (72B) Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH25%
71Llama 3.2 11B Vision Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH19.3%
72Llama 3.2 11B Vision Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH18.8%
73Qwen2 VL 7b Instructsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH17.3%
74Qwen2 VL 7b Instructsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH16.9%
75LFM-7bsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH16.7%
76LFM-7bsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH16.2%
77Pixtral 12Bsequences_series, text, gpt-4o-2024-08-06· U-MATH and mu-MATH13.3%
78Pixtral 12Bsequences_series, all, gpt-4o-2024-08-06· U-MATH and mu-MATH13%
79Claude 3.5 Sonnetsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
80GPT 4osequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
81LFM-7bsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
82Llama 3.1 8B Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
83Llama 3.2 11B Vision Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
84Nvidia Llama 3.1 Nemotron 70B Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
85Phi 4sequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
86Pixtral 12Bsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
87QVQ 72B Previewsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
88Qwen2 VL 7b Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
89Qwen2.5 32B Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%
90Qwen2.5 7B Instructsequences_series, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH0%

Showing 51–90 of 90 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)