Grade School Math 8K scores from Epoch AI's data collection. The same benchmark as the standard GSM8K — 8,500 grade-school math word problems.
This definition has not completed editorial and provenance review. Verify its metric, version, source, and evaluation configuration upstream before using it for a model-selection decision.
Well-established math reasoning baseline with extensive historical data.
Non-standard score representation in this dataset (range 0–8, not 0–100%). Saturated benchmark.
Higher is better. Score interpretation varies (Epoch-specific encoding).
Non-standard score representation in this dataset (range 0–8, not 0–100%).
| # | Model | Score |
|---|---|---|
| 1 | 92.0 | |
| 2 | 72.8 | |
| 3 | 68.9 | |
| 4 | 57.4 | |
| 5 | 47.4 | |
| 6 | 45.8 | |
| 7 | 45.6 | |
| 8 | 41.2 | |
| 9 | 41.0 | |
| 10 | 40.5 | |
| 11 | 40.4 | |
| 12 | 40.1 | |
| 13 | 39.4 | |
| 14 | 38.8 | |
| 15 | 36.2 | |
| 16 | 34.0 | |
| 17 | 33.7 | |
| 18 | 33.2 | |
| 19 | 32.3 | |
| 20 | 31.9 | |
| 21 | 31.4 | |
| 22 | 28.5 | |
| 23 | 27.8 | |
| 24 | 27.7 | |
| 25 | 27.1 | |
| 26 | 26.8 | |
| 27 | 26.8 | |
| 28 | 26.4 | |
| 29 | Gemini 2.5 Flash-Lite Preview (Sep '25) (Non-reasoning) source benchpress Evidence observed Apr 13, 2026 | 25.2 |
| 30 | 25.1 | |
| 31 | 23.4 | |
| 32 | 22.4 | |
| 33 | 21.8 | |
| 34 | 21.3 | |
| 35 | 20.7 | |
| 36 | 20.3 | |
| 37 | 19.1 | |
| 38 | 18.7 | |
| 39 | 18.7 | |
| 40 | 18.3 | |
| 41 | 18.2 | |
| 42 | 17.9 | |
| 43 | 17.8 | |
| 44 | 17.3 | |
| 45 | 16.9 | |
| 46 | 16.6 | |
| 47 | 16.3 | |
| 48 | 16.0 | |
| 49 | 16.0 | |
| 50 | 15.0 |