Jina OCR vs dots.ocr
Pick Jina OCR for non-commercial PDF to Markdown work, research and evaluation, where its 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench help. Pick dots.ocr when you need a commercially usable MIT-licensed model that returns layout boxes, categories and text in one pass. The main split is licence: Jina OCR needs a separate license for commercial use, dots.ocr does not.
Jina OCRjina/jina-ocr | dots.ocrrednote/dots-ocr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Clean Markdown at low cost | Multilingual document parsing |
| Inputs | PDF, images | PDF, images |
| Languages | About 100 languages | 100+ languages |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| 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 is available as a hosted API as well as self-hosted weights, so you can test without running a GPU.
- It scores 83.4 on olmOCR-Bench, 7.4 points above DeepSeek-OCR, with only about 570M active parameters.
- It runs at 2.57 pages per second on an A100, and falls back to CPU in Transformers if you have no GPU.
Choose dots.ocr if
- Its code is MIT-licensed, so you are not blocked from commercial use by a non-commercial clause.
- One pass returns layout bounding boxes, categories and text, with tables as HTML and formulas as LaTeX.
- The dots.mocr-svg variant converts charts and diagrams to SVG, and dots.mocr scores 83.9 on olmOCR-bench.
Key differences
| Jina OCR | dots.ocr | |
|---|---|---|
| Licence | CC-BY-NC-4.0; commercial use needs a license from Jina/Elastic | MIT, plus a separate dots.ocr license agreement file in the repository |
| How you run it | Self-host or Jina's hosted API | Self-host only, via vLLM or Hugging Face Transformers |
| Size | 3B MoE, about 570M active | 3B, with a 1.7B LLM backbone |
| Output formats | Markdown, HTML, LaTeX | JSON, Markdown, HTML, LaTeX, bounding boxes, SVG |
| olmOCR-bench | 83.4 (authors) | 79.1 for dots.ocr, 83.9 for dots.mocr 1.5 (authors) |
| Hardware | GPU recommended (A100 and L4 benchmarks); vLLM 0.21+ for FastMTP; CPU fallback in Transformers | Single CUDA GPU with vLLM is the documented path; no VRAM figure stated |
| Known failure | Hosted API returns HTTP 503 on cold start; retry after 30 to 60 seconds | Runs of ellipses or underscores can make output repeat endlessly; a table above 1000 grid cells drops the whole page in Docling |
Sources
- jina-ocr-v1 model card
- Jina model page
- dots.ocr GitHub README
- dots.ocr blog / limitations
- Docling issue #5146
What each one is
Jina OCR
Compact page-to-Markdown model from Jina AI, built on the DeepSeek-OCR encoder.
Jina OCR detailsdots.ocr
Single vision model from rednote that does layout detection and content recognition together.
dots.ocr 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": ["rednote/dots-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Jina OCR alternativesdots.ocr alternativesBest OCR models in 2026