SEA-Vision - Document Parsing: leaderboard
Metric: Normalized edit distance (0-1) between the parsed page and the annotation, mean of the 11 language averages, on SEA-Vision end-to-end document parsing (15,234 pages of nine document types across 11 Southeast Asian languages, parsed to structured text, tables, formulas and reading order and scored OmniDocBench-style); lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 0.16 | #145 |
| 2 | Qwen 3 VL 32B Instruct | 0.23 | #276 |
| 3 | Qwen 2.5 VL 72B Instruct | 0.27 | #364 |
| 4 | GPT-4o | 0.31 | #333 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=sea-vision-document-parsing · How It Works · Data refreshed daily, snapshot 2026-10-11.