Easy subset of LongBench v2 — long-context questions where the answer can be found with relatively straightforward retrieval from the provided long document.
Tests basic long-context retrieval without requiring complex reasoning.
Easier tasks provide less differentiation among frontier models.
Higher is better. Accuracy percentage (0–100%).
Easier tasks provide less differentiation among frontier models.
| # | Model | Score |
|---|---|---|
| 1 | 81.8 | |
| 2 | 80.4 | |
| 3 | 80.2 | |
| 4 | 79.4 | |
| 5 | 78.8 | |
| 6 | 77.8 | |
| 7 | 75.2 | |
| 8 | 74.8 | |
| 9 | 74.4 | |
| 10 | 73.6 | |
| 11 | 72.9 | |
| 12 | 72.7 | |
| 13 | 72.3 | |
| 14 | 71.5 | |
| 15 | 71.3 | |
| 16 | 70.8 | |
| 17 | 70.5 | |
| 18 | 70.3 | |
| 19 | 70.0 | |
| 20 | 69.9 | |
| 21 | 68.9 | |
| 22 | 68.4 | |
| 23 | 67.3 | |
| 24 | 67.2 | |
| 25 | 67.1 | |
| 26 | 67.0 | |
| 27 | 66.9 | |
| 28 | 66.5 | |
| 29 | 66.5 | |
| 30 | 65.8 | |
| 31 | 64.5 | |
| 32 | 63.3 | |
| 33 | 63.2 | |
| 34 | 62.7 | |
| 35 | 62.5 | |
| 36 | 62.2 | |
| 37 | 62.1 | |
| 38 | 62.1 | |
| 39 | 61.7 | |
| 40 | 61.5 | |
| 41 | 61.3 | |
| 42 | 61.0 | |
| 43 | 60.5 | |
| 44 | 60.4 | |
| 45 | 60.4 | |
| 46 | 60.3 | |
| 47 | 59.9 | |
| 48 | 59.7 | |
| 49 | 59.5 | |
| 50 | 59.2 |