Enterprise document processing sub-score from the IDP Leaderboard testing key information extraction, OCR, table parsing, and visual QA across 6,406 samples — the tasks enterprises most commonly need.
Most directly enterprise-relevant; tests extraction + VQA together; uses real-world document types.
Fewer samples than OlmOCR and OmniDoc sub-benchmarks; composite across diverse task types.
Higher is better. Composite score across KIE, OCR, table parsing, and VQA tasks, 0-100.
Max score: 100
Fewer samples than OlmOCR and OmniDoc sub-benchmarks.
| # | Model | Score |
|---|---|---|
| 1 | 89.6 | |
| 2 | 84.4 | |
| 3 | 84.4 | |
| 4 | 81.8 | |
| 5 | 81.2 | |
| 6 | 81.1 | |
| 7 | 80.5 | |
| 8 | 80.5 | |
| 9 | 80.2 | |
| 10 | 80.0 | |
| 11 | 77.4 | |
| 12 | 76.2 | |
| 13 | 74.7 | |
| 14 | 74.5 | |
| 15 | 73.8 | |
| 16 | 73.3 | |
| 17 | 72.9 | |
| 18 | 72.9 | |
| 19 | 71.7 | |
| 20 | 68.5 | |
| 21 | 67.1 | |
| 22 | 65.8 | |
| 23 | 61.2 | |
| 24 | 59.0 | |
| 25 | 55.0 | |
| 26 | 54.9 | |
| 27 | 44.1 |