Docling vs Kraken
Pick Docling to ingest born-digital PDFs, Office files and HTML into structured JSON for RAG pipelines. Pick Kraken for historical manuscripts, early prints and non-Latin or right-to-left scripts, where you can train the layout and recognition models on your own material. The main split is document type: Docling targets modern files, Kraken targets historical and non-Latin material.
Doclingibm/docling | Krakenkraken/kraken | |
|---|---|---|
| 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| Best for | Structured export for RAG | Historical documents and manuscripts |
| Inputs | PDF, images, DOCX | images |
| Languages | Latin scripts, CJK | Many scripts, incl. right-to-left |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | ✓ Yes |
| 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 Docling if
- It reads PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL through one API.
- It exports lossless JSON that keeps layout, reading order and table structure, plus Markdown, HTML, DocTags and plain text.
- It runs on macOS, Linux and Windows, on CPU or GPU, and has LangChain and LlamaIndex integrations.
Choose Kraken if
- It is built for historical documents and non-Latin scripts, with right-to-left, BiDi and top-to-bottom support.
- Layout analysis, reading order and character recognition are all trainable on your own material.
- It outputs ALTO, PageXML, abbyyXML and hOCR for digital humanities tools, and community-trained models are on Zenodo.
Key differences
| Docling | Kraken | |
|---|---|---|
| Best document types | Born-digital PDFs and Office files | Historical manuscripts and early prints |
| Licence | MIT (individual models have their own licences) | Apache-2.0 |
| Platforms | macOS, Linux and Windows, x86_64 and arm64 | Linux and macOS only, x64 or ARM, Python 3.10 to 3.13. No Windows support. |
| Output formats | Markdown, HTML, JSON, DocTags, plain text | ALTO, PageXML, hOCR, plain text, bounding boxes |
| OCR approach | Plugs in EasyOCR, Tesseract, RapidOCR or a VLM such as the 258M GraniteDocling. Scan accuracy depends on the engine you choose. | Its own trainable recognition. It usually needs training on your own material for good accuracy. |
| 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). | Install can fail with a dependency conflict (issue #816). Learned line order is lost in segmentation and text output (issue #820). |
| Benchmarks | Not published | Not published |
Sources
- Docling GitHub README
- Supported formats
- Issue #5254 OCR regression
- Kraken GitHub README
- Zenodo OCR models
- Issue #820
What each one is
Kraken
OCR engine for historical documents and manuscripts, with trainable models for unusual scripts.
Kraken 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": ["kraken/kraken"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling vs TrOCRDocling vs docTRTesseract vs KrakenMinerU vs KrakenLlamaParse vs DoclingPaddleOCR-VL vs DoclingMinerU vs DoclingDocling vs Tesseract
Docling alternativesKraken alternativesBest OCR models in 2026