Jina OCR vs Kraken
Pick Jina OCR to turn modern PDF pages into Markdown, HTML or LaTeX, but only for non-commercial work, because its CC-BY-NC-4.0 licence blocks commercial use without a separate deal. Pick Kraken for historical manuscripts, early prints and non-Latin or right-to-left scripts, where you can train layout and recognition on your own material under Apache-2.0. They do different jobs: Jina OCR is a ready page-to-Markdown model, and Kraken is a trainable toolkit for archival text.
Jina OCRjina/jina-ocr | Krakenkraken/kraken | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 1 credit / page |
| Per 1,000 pages | $1.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Clean Markdown at low cost | Historical documents and manuscripts |
| Inputs | PDF, images | images |
| Languages | About 100 languages | Many scripts, incl. right-to-left |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | ✓ Yes |
| 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 Jina OCR if
- It scored 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench (authors' figures), 7.4 points above DeepSeek-OCR on olmOCR-Bench.
- It has only about 570M active parameters (3B MoE) and runs at 2.57 pages per second on an A100.
- You can self-host it or call Jina's hosted API, and it outputs Markdown, HTML and LaTeX.
Choose Kraken if
- Its licence is Apache-2.0, with no commercial restriction.
- It is built for historical documents and non-Latin scripts, with right-to-left, BiDi and top-to-bottom support.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, and a public Zenodo repository holds community-trained models.
Key differences
| Jina OCR | Kraken | |
|---|---|---|
| Licence | CC-BY-NC-4.0; commercial use needs a licence from Jina/Elastic | Apache-2.0 |
| How you run it | Self-host the open weights or use Jina's hosted API | Local Python library installed with pip; no hosted option |
| Platform and hardware | GPU recommended (A100 and L4 benchmarked); vLLM 0.21+ for FastMTP; CPU fallback in Transformers | Linux or macOS only, x64 or ARM, Python 3.10 to 3.13; no Windows |
| Output formats | Markdown, HTML, LaTeX | ALTO, PageXML, hOCR, plain text, bounding boxes |
| Benchmarks | OmniDocBench v1.6 overall 91.14; olmOCR-Bench overall 83.4 | Not published |
| Best document types | Non-commercial PDF to Markdown, research and evaluation | Historical manuscripts, early prints, Arabic, Hebrew and other RTL scripts, digital humanities projects |
| Known failure | Hosted API returns HTTP 503 on cold start; retry after 30 to 60 seconds. Reported scores use a stricter prompt than the recommended default | Install can fail with a dependency conflict (issue #816); learned line order is lost in segmentation and text output (issue #820); usually needs training on your own material |
Sources
What each one is
Jina OCR
Compact page-to-Markdown model from Jina AI, built on the DeepSeek-OCR encoder.
Jina OCR detailsKraken
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.
Use one, fall back to the other
{"model": "jina/jina-ocr","fallbacks": ["kraken/kraken"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Jina OCR alternativesKraken alternativesBest OCR models in 2026