Benchmark
U-MATH overall 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
- 1.61
Score distribution
- 44.8%
- Mean
- 43.9%
- Median
- 22.3%
- Std dev
- 16%
- 10th percentile
- 80.7%
- 90th percentile
- 9%
- Min
This benchmark has a weight of 1.61 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
| # | Model | Score |
|---|---|---|
| 1 | o1u_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 93.1% |
| 2 | o3 Miniu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 92.8% |
| 3 | DeepSeek R1u_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 91.3% |
| 4 | Gemini 2.0 Flash Experimentalu_math, text, gpt-4o-2024-08-06, thinking· U-MATH and mu-MATH | 89.2% |
| 5 | o1u_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 86.8% |
| 6 | Gemini 2.0 Flash Experimentalu_math, all, gpt-4o-2024-08-06, thinking· U-MATH and mu-MATH | 83.6% |
| 7 | O1 Miniu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 82.9% |
| 8 | QwQ 32B Previewu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 82.7% |
| 9 | o3 Miniu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 82.2% |
| 10 | DeepSeek R1u_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 80.7% |
| 11 | O1 Miniu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 76.3% |
| 12 | QwQ 32B Previewu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 73.1% |
| 13 | QVQ 72B Previewu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 69.7% |
| 14 | DeepSeek V3u_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 69.3% |
| 15 | Qwen2.5 Math 72B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 68.7% |
| 16 | QVQ 72B Previewu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 65% |
| 17 | Gemini 1.5 Prou_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 63.4% |
| 18 | DeepSeek V3u_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 62.6% |
| 19 | Gemini 1.5 Flashu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 61.2% |
| 20 | Qwen2.5 32B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 60.4% |
| 21 | Gemini 1.5 Prou_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 60.1% |
| 22 | Qwen2.5 Math 72B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 59.5% |
| 23 | Qwen2.5 72B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 58.9% |
| 24 | Gemini 2.0 Flash Experimentalu_math, visual, gpt-4o-2024-08-06, thinking· U-MATH and mu-MATH | 58.5% |
| 25 | o1u_math, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH | 58.5% |
| 26 | Phi 4u_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 58.3% |
| 27 | Gemini 1.5 Flashu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 57.8% |
| 28 | Mistral Large Instruct 2411u_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 55.6% |
| 29 | Phi 4u_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 54.5% |
| 30 | GPT 4ou_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 53.9% |
| 31 | Qwen2.5 Math 7B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 53% |
| 32 | Qwen2.5 32B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 52.4% |
| 33 | Llama 3.3 70B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 51.7% |
| 34 | Qwen2.5 72B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 51.2% |
| 35 | Qwen2.5 7B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 50.4% |
| 36 | GPT 4ou_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 50.2% |
| 37 | Nvidia Llama 3.1 Nemotron 70B Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 47.7% |
| 38 | Mistral Large Instruct 2411u_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 47.6% |
| 39 | GPT-4o miniu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 47.2% |
| 40 | O1 Miniu_math, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH | 46.5% |
| 41 | Qwen2.5 Math 7B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 45.5% |
| 42 | Gemini 1.5 Prou_math, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH | 45% |
| 43 | Llama 3.3 70B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 44.7% |
| 44 | QVQ 72B Previewu_math, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH | 44% |
| 45 | Qwen2-VL (72B) Instructu_math, text, gpt-4o-2024-08-06· U-MATH and mu-MATH | 43.9% |
| 46 | GPT-4o miniu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 43.4% |
| 47 | Qwen2.5 7B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 43.3% |
| 48 | Nvidia Llama 3.1 Nemotron 70B Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 42.5% |
| 49 | Gemini 1.5 Flashu_math, visual, gpt-4o-2024-08-06· U-MATH and mu-MATH | 42.5% |
| 50 | Qwen2-VL (72B) Instructu_math, all, gpt-4o-2024-08-06· U-MATH and mu-MATH | 41.8% |
Showing 1–50 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)