DROP: leaderboard
Discrete Reasoning Over Paragraphs: reading comprehension requiring discrete operations like addition, counting, and sorting over text passages.
Metric: F1 Score. Source: github.com. Saturation forecast: Estimated already saturated. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 | 90.2 | #183 |
| 2 | O3 (High) | 89.8 | #124 (O3) |
| 3 | Claude 3.5 Sonnet | 87.1 | #350 |
| 4 | GPT-4 Turbo | 86 | #404 |
| 5 | Llama 3.1 405B | 84.8 | #466 |
| 6 | O1 Mini | 83.9 | #323 |
| 7 | GPT-4o (2024-05-13) | 83.7 | #325 |
| 8 | GPT-4.5 (Preview) | 83.4 | #244 |
| 9 | GPT-4 Preview (1106) | 83.2 | #418 |
| 10 | Claude 3 Opus | 83.1 | #571 |
| 11 | O3 (Low) | 82.3 | #124 (O3) |
| 12 | GPT-4.1 Nano | 82.2 | #731 |
| 13 | GPT-4o (2024-11-20) | 81.5 | #381 |
| 14 | GPT-4 Preview (0125) | 81.5 | #486 |
| 15 | GPT-4.1 Mini | 81 | #355 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=drop · How It Works · Data refreshed daily, snapshot 2026-10-20.