Docling vs Granite Docling
Pick Docling when you need one library that ingests PDF, DOCX, PPTX, XLSX, HTML and EPUB and exports lossless JSON for RAG. Pick Granite Docling when you want a small 258M VLM for English PDFs with tables, code and equations, usually run inside Docling. They are not rivals: Docling is the pipeline and Granite Docling is one model you can plug into it.
Doclingibm/docling | Granite Doclingibm/granite-docling | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 2.2 s / page | 1.5 s / page |
| Best for | Structured export for RAG | Small-model document conversion |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | Latin scripts, CJK | English, experimental CJK and Arabic |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | No |
| 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 Docling if
- You need to parse many formats in one API, including PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL.
- You want lossless JSON export that keeps layout, reading order and table structure for downstream tools, or LangChain and LlamaIndex integrations.
- You want to choose the OCR backend yourself, such as EasyOCR, Tesseract or RapidOCR, or swap in a VLM.
Choose Granite Docling if
- You convert English PDFs with tables, code and equations and want a model that reports table structure TEDS 0.97 on FinTabNet and equation F1 0.968.
- You need low-cost conversion on CPU or a small GPU, since the model has only 258M parameters.
- You want an Apache-2.0 model that outputs DocTags, Markdown, HTML, LaTeX and bounding boxes.
Key differences
| Docling | Granite Docling | |
|---|---|---|
| What it is | Python library that converts PDFs, Office files, HTML and images into a DoclingDocument | 258M document conversion VLM from IBM that outputs DocTags, usually run inside Docling |
| Licence | MIT (individual models have their own licences) | Apache-2.0 |
| Price and hosting | Free, self-hosted; you pay for compute | Free, self-hosted; you pay for compute |
| Output formats | Markdown, HTML, JSON, DocTags, plain text | DocTags, Markdown, HTML, LaTeX, bounding boxes |
| Benchmarks | Not published | Table TEDS 0.97 (FinTabNet, 150dpi), full-page OCR F1 0.84, equation F1 0.968, OCRBench 500 (all from the authors) |
| Languages | Depends on the OCR engine you plug in | English primary; Japanese, Arabic and Chinese marked experimental |
| Known failure | Version 2.143.1 drops words and numbers with full-page Tesseract OCR (issue #5254); reading order can merge neighbouring columns (issue #5132) | Can get stuck in repetitive loops on page headings and invoices; on GPUs without bfloat16 such as T4 may output only exclamation marks unless run in float32 |
Sources
- Docling GitHub README
- Supported formats
- Granite-Docling model card
- Issue #5254 OCR regression
- HF discussion #24 repetition
What each one is
Granite Docling
Compact IBM vision model that converts pages to DocTags for the Docling pipeline.
Granite Docling detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "ibm/docling","fallbacks": ["ibm/granite-docling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling vs SmolDoclingDocling vs Qwen-VLTesseract vs Granite DoclingMinerU vs Granite DoclingLlamaParse vs DoclingPaddleOCR-VL vs DoclingMinerU vs DoclingDocling vs Tesseract
Docling alternativesGranite Docling alternativesBest OCR models in 2026