Qianfan OCR vs Unstructured
Pick Qianfan OCR when you want a self-hosted model that parses scans, handwriting and multilingual documents into Markdown, HTML or LaTeX. Pick Unstructured when you need to ingest many file types (PDF, Word, PowerPoint, email, HTML) into a RAG pipeline. The main split is scope: Qianfan OCR is a single 4B document model, while Unstructured is an ingestion library and API that also covers non-image formats.
Qianfan OCRbaidu/qianfan-ocr | Unstructuredunstructured/unstructured | |
|---|---|---|
| Type | OCR API | OCR API |
| Price | 2 credits / page | 20 credits / page |
| Per 1,000 pages | $2.00 | $20.00 |
| Typical speed | 2.0 s / page | 3.0 s / page |
| Best for | Chinese documents | Mixed file types into RAG |
| Inputs | PDF, images | PDF, images, DOCX |
| Languages | Chinese, English and more | Multilingual |
| Max pages per request | 200 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | No |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Qianfan OCR if
- It scores 93.12 on OmniDocBench v1.5, the top end-to-end score in Baidu's own table.
- It covers 192 languages plus handwriting and scene text, while Unstructured table OCR ignores language settings.
- Its Apache-2.0 weights are free to self-host for commercial use, and it outputs Markdown, JSON, HTML, LaTeX and bounding boxes.
Choose Unstructured if
- It handles PDFs, HTML, Word, PowerPoint, emails and images in one library, with many connectors.
- The hosted API gives you 10,000 free pages, then $0.015 per page.
- Its Fast, Auto, High-Res and VLM strategies let you trade speed for quality, and Fast suits born-digital PDFs.
Key differences
| Qianfan OCR | Unstructured | |
|---|---|---|
| What it is | 4B end-to-end document parsing model from Baidu (Qwen3-4B language model) | Document ETL library and hosted platform that splits files into typed elements |
| Licence | Apache-2.0 | Apache-2.0 for the open-source library; proprietary for the hosted platform |
| How you run it | Self-host the weights (benchmarked on one A100 with vLLM) or use the Baidu Qianfan platform | Self-host the Python library or call the hosted API |
| Price | Free self-hosted, you pay for compute; hosted pricing on Qianfan not published | Library free; hosted first 10,000 pages free, then $0.015 per page; Business is custom |
| Output formats | Markdown, JSON, HTML, LaTeX, bounding boxes | JSON, HTML, plain text, bounding boxes |
| Tables | Listed as a good fit for tables and formulas | Open-source table cell content scores 0.426 versus 0.866 hosted (vendor benchmark); hi_res tables can cut off at edges and drop first letters |
| Known limits | 1.024 pages per second on an A100 even with W8A8; thinking mode can emit up to 16,384 tokens; not for CPU-only deployment | Table OCR ignores language settings; extra Tesseract pass on tables adds time; library sends usage telemetry by default |
Sources
- Qianfan-OCR model card
- Qianfan-OCR paper
- Unstructured pricing
- Unstructured GitHub README
- Unstructured issue #1694
- Unstructured issue #3262
What each one is
Qianfan OCR
Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.
Qianfan OCR detailsUnstructured
Document ETL for RAG. Partitions many file types into typed elements; open-source library too.
Unstructured detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "baidu/qianfan-ocr","fallbacks": ["unstructured/unstructured"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qianfan OCR alternativesUnstructured alternativesBest OCR models in 2026