SciDocBench - Information Extraction: leaderboard
Metric: Score (%). Source: arxiv.org. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.8 | 57.12 |
| 2 | GPT-5.6 Sol | 56.78 |
| 3 | Claude Opus 5 | 55.52 |
| 4 | GPT-5.6 Terra | 51.74 |
| 5 | Claude Sonnet 4.6 | 51.51 |
| 6 | Kimi K2.5 | 50.62 |
| 7 | GPT-5.6 Luna | 47.52 |
| 8 | Claude Haiku 4.5 | 42.83 |
Interactive version: theaggregate.ai/benchmark?slug=scidocbench-information-extraction · How It Works · Data refreshed daily, snapshot 2026-09-20.