Benchmark
mu-MATH PPV
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
- 1.50
Score distribution
- 77.7%
- Mean
- 81.2%
- Median
- 11.5%
- Std dev
- 62.8%
- 10th percentile
- 89.7%
- 90th percentile
- 35.5%
- Min
This benchmark has a weight of 1.50 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
| # | Model | Score |
|---|---|---|
| 51 | Qwen2.5 Math 7B InstructGemini-1.5-Pro, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 78.2% |
| 52 | Qwen2.5 72B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 77.5% |
| 53 | Mistral Large Instruct 2411GPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 77.4% |
| 54 | GPT-4o miniLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 77.3% |
| 55 | Llama 3.1 8B InstructGemini-1.5-Pro, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 77% |
| 56 | Gemini 2.0 Flash Experimentalmu_math, PPV, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH | 76.9% |
| 57 | QwQ 32B PreviewGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 75.6% |
| 58 | Gemini 2.0 Flash ExperimentalQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH | 75% |
| 59 | Qwen2.5 72B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 74.2% |
| 60 | Gemini 2.0 Flash ExperimentalLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH | 73.9% |
| 61 | Qwen2.5 72B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 73.8% |
| 62 | Qwen2.5 Math 72B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 73.8% |
| 63 | Qwen2.5 72B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 72.3% |
| 64 | Llama 3.1 70B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 71.8% |
| 65 | Qwen2.5 Math 72B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 71.6% |
| 66 | Mistral Large Instruct 2411Llama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 70.9% |
| 67 | QwQ 32B PreviewLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 70.4% |
| 68 | Gemini 2.0 Flash ExperimentalGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct, thinking· U-MATH and mu-MATH | 70.2% |
| 69 | Qwen2.5 7B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 69.3% |
| 70 | Qwen2.5 Math 72B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 66.9% |
| 71 | Qwen2.5 7B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 66.5% |
| 72 | Qwen2.5 7B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 66.3% |
| 73 | Qwen2.5 Math 72B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 65.8% |
| 74 | Qwen2.5 Math 7B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 64.7% |
| 75 | Llama 3.1 70B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 64.1% |
| 76 | Qwen2.5 Math 7B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 62.9% |
| 77 | Llama 3.1 70B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 62.8% |
| 78 | Qwen2.5 7B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 59.4% |
| 79 | Qwen2.5 Math 7B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 57.7% |
| 80 | Llama 3.1 8B InstructQwen2.5-72B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 56.5% |
| 81 | Llama 3.1 8B Instructmu_math, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 56% |
| 82 | Llama 3.1 8B InstructGPT-4o, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 53.3% |
| 83 | Qwen2.5 Math 7B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 49.7% |
| 84 | Llama 3.1 70B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 43.7% |
| 85 | Llama 3.1 8B InstructLlama-3.1-70B-Instruct, PPV, Qwen/Qwen2.5-72B-Instruct· U-MATH and mu-MATH | 35.5% |
Showing 51–85 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)