README.md

---
title: OCR Demo
emoji: ⚡
colorFrom: indigo
colorTo: purple
sdk: static
pinned: false
app_file: index.html
short_description: Unified 7-Model OCR Benchmark & Layout Visualizer
---

# ⚡ Unified 7-Model OCR Benchmark & Layout Visualizer

An automated benchmark and spatial layout evaluation suite for 7 state-of-the-art OCR & Document AI models:

1. **DeepSeek-OCR** (`deepseek-ai/DeepSeek-OCR`)
2. **Qwen3-VL** (`Qwen/Qwen2.5-VL-3B-Instruct`)
3. **PP-StructureV3** (`PaddleOCR/PP-StructureV3`)
4. **NuExtract3** (`numind/NuExtract-v1.5`)
5. **PP-OCRv5** (`PaddleOCR/PP-OCRv5`)
6. **olmOCR** (`allenai/olmOCR-7B-0225-preview`)
7. **MinerU 3** (`opendatalab/MinerU2.5-Pro-2605-1.2B`)