Granite Docling vs Unstructured
Pick Granite Docling to convert English PDFs with tables, code and equations on your own hardware, since it is a 258M model with a 0.97 table structure score on FinTabNet. Pick Unstructured to ingest mixed files such as Word, PowerPoint, HTML and email for RAG, especially if you want a hosted API with 10,000 free pages. Table accuracy is the main split: the open-source Unstructured library scores 0.426 on table cell content versus 0.866 hosted.
Granite Doclingibm/granite-docling | Unstructuredunstructured/unstructured | |
|---|---|---|
| Type | Open source | OCR API |
| Price | 1 credit / page | 20 credits / page |
| Per 1,000 pages | $1.00 | $20.00 |
| Typical speed | 1.5 s / page | 3.0 s / page |
| Best for | Small-model document conversion | Mixed file types into RAG |
| Inputs | PDF, images | PDF, images, DOCX |
| Languages | English, experimental CJK and Arabic | Multilingual |
| Max pages per request | 500 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | ✓ 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 Granite Docling if
- You convert English PDFs with tables, code and equations, where it reports table structure TEDS 0.97 on FinTabNet, equation F1 0.968 and code F1 0.988.
- You want low-cost conversion on CPU or a small GPU, because the model has only 258M parameters and is Apache-2.0 licensed.
- You already use Docling and want output as DocTags, Markdown, HTML, LaTeX or bounding boxes.
Choose Unstructured if
- You ingest many file types for RAG, since one library handles PDFs, HTML, Word, PowerPoint, emails and images.
- You want a hosted API with no setup, with the first 10,000 pages free and then $0.015 per page.
- You process born-digital PDFs and can use the fast strategy, or pick among Auto, Fast, High-Res and VLM to trade speed for quality.
Key differences
| Granite Docling | Unstructured | |
|---|---|---|
| Licence | Apache-2.0 | Apache-2.0 for the open-source library; proprietary for the hosted platform |
| How you run it | Self-hosted only, typically inside Docling; you pay for compute | Self-host the library or call the hosted API |
| Price | Free, self-hosted | Hosted: first 10,000 pages free, then $0.015 per page; Business is custom. Library is free |
| Input scope | Document conversion of English PDFs with tables, code and equations | PDFs, HTML, Word, PowerPoint, emails and images |
| Table accuracy | TEDS structure 0.97 on FinTabNet (150dpi) | Table cell content 0.426 open-source versus 0.866 hosted (vendor benchmark) |
| Languages | English primary; Japanese, Arabic and Chinese are experimental | Table OCR ignores language settings (issue #1694) |
| Known failure | Can loop on page headings and invoices; may print only exclamation marks on GPUs without bfloat16, such as T4, unless run in float32 | hi_res tables can cut off at the edges and drop first letters (issue #3262); library sends usage telemetry by default |
Sources
- Granite-Docling model card
- HF discussion #24 repetition
- Unstructured pricing
- Unstructured GitHub README
- Issue #3262
- Issue #1694
What each one is
Granite Docling
Compact IBM vision model that converts pages to DocTags for the Docling pipeline.
Granite Docling 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": "ibm/granite-docling","fallbacks": ["unstructured/unstructured"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Granite Docling alternativesUnstructured alternativesBest OCR models in 2026