GAIA Level 2 tasks — medium difficulty requiring multi-step tool use, information synthesis across sources, and moderate reasoning chains.
Good discriminator for agentic capability. Tasks require planning and multi-step execution.
Wide score variance across models. Difficulty calibration can shift between evaluation rounds.
Higher is better. Accuracy percentage (0–100%).
Wide score variance across models.
| # | Model | Score |
|---|---|---|
| 1 | 96.6 | |
| 2 | 96.5 | |
| 3 | 95.6 | |
| 4 | 91.7 | |
| 5 | 90.9 | |
| 6 | 90.7 | |
| 7 | 90.7 | |
| 8 | 88.7 | |
| 9 | 88.5 | |
| 10 | 86.4 | |
| 11 | 85.0 | |
| 12 | 84.8 | |
| 13 | 84.5 | |
| 14 | 83.8 | |
| 15 | 82.7 | |
| 16 | 82.2 | |
| 17 | 82.1 | |
| 18 | 82.0 | |
| 19 | 81.0 | |
| 20 | 80.6 | |
| 21 | 80.3 | |
| 22 | 80.2 | |
| 23 | 80.2 | |
| 24 | 79.6 | |
| 25 | 79.0 | |
| 26 | 78.8 | |
| 27 | 77.8 | |
| 28 | 77.5 | |
| 29 | 77.3 | |
| 30 | 76.7 | |
| 31 | 76.3 | |
| 32 | 76.2 | |
| 33 | 76.1 | |
| 34 | 76.0 | |
| 35 | 75.6 | |
| 36 | 75.6 | |
| 37 | 75.5 | |
| 38 | 75.4 | |
| 39 | 75.3 | |
| 40 | 74.8 | |
| 41 | 74.6 | |
| 42 | 74.4 | |
| 43 | 73.1 | |
| 44 | 72.7 | |
| 45 | 71.8 | |
| 46 | 71.8 | |
| 47 | 71.3 | |
| 48 | 71.2 | |
| 49 | 70.9 | |
| 50 | 70.3 |