Grade School Math 8K — 8,500 grade-school level math word problems requiring multi-step arithmetic reasoning. Created by OpenAI.
Well-established baseline for mathematical reasoning. Multi-step problems test chain-of-thought ability.
Now largely saturated — frontier models score 95–99%. Heavy contamination risk as it's been in training data for years.
Higher is better. Accuracy percentage (0–100%). Frontier models score >95%.
Now largely saturated — frontier models score 95–99%.
| # | Model | Score |
|---|---|---|
| 1 | 99.0 | |
| 2 | 99.0 | |
| 3 | 98.7 | |
| 4 | 98.0 | |
| 5 | 97.9 | |
| 6 | 97.9 | |
| 7 | 97.9 | |
| 8 | 97.5 | |
| 9 | 97.5 | |
| 10 | 97.4 | |
| 11 | 97.3 | |
| 12 | 97.3 | |
| 13 | 97.3 | |
| 14 | 97.2 | |
| 15 | 97.2 | |
| 16 | 97.1 | |
| 17 | 97.1 | |
| 18 | 97.1 | |
| 19 | 97.1 | |
| 20 | 97.0 | |
| 21 | 97.0 | |
| 22 | 96.9 | |
| 23 | 96.9 | |
| 24 | 96.9 | |
| 25 | 96.9 | |
| 26 | 96.8 | |
| 27 | 96.8 | |
| 28 | 96.7 | |
| 29 | 96.7 | |
| 30 | 96.7 | |
| 31 | 96.7 | |
| 32 | 96.6 | |
| 33 | 96.6 | |
| 34 | 96.6 | |
| 35 | 96.5 | |
| 36 | 96.5 | |
| 37 | 96.5 | |
| 38 | 96.4 | |
| 39 | 96.4 | |
| 40 | 96.3 | |
| 41 | 96.3 | |
| 42 | 96.3 | |
| 43 | 96.3 | |
| 44 | 96.2 | |
| 45 | 96.2 | |
| 46 | 96.2 | |
| 47 | 96.2 | |
| 48 | 96.2 | |
| 49 | 96.2 | |
| 50 | 96.1 |