Jina OCR vs TrOCR
Pick Jina OCR when you need to turn whole PDF pages into Markdown, HTML or LaTeX, and your use is non-commercial or you have a license from Jina/Elastic. Pick TrOCR when you already have cropped single text-line images, especially English handwriting, and want an MIT-licensed recognizer you can fine-tune. The main split is scope: Jina OCR reads pages, TrOCR reads one line at a time.
Jina OCRjina/jina-ocr | TrOCRmicrosoft/trocr | |
|---|---|---|
| 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 | 1.2 s / page |
| Best for | Clean Markdown at low cost | Handwritten text lines |
| Inputs | PDF, images | images |
| Languages | About 100 languages | English |
| Max pages per request | 200 | 100 |
| Tables | ✓ Yes | No |
| Handwriting | No | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | No | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | ✓ Yes | 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 Jina OCR if
- It converts full pages to Markdown, HTML or LaTeX, while TrOCR only returns plain text for a single line.
- It scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, per the authors.
- You can self-host it or call Jina's hosted API, and it has about 570M active parameters out of 3B (MoE).
Choose TrOCR if
- It is MIT licensed, so you can use it in commercial products without a separate license.
- It reports a 2.89 character error rate on the IAM handwriting set with the large model, per the paper.
- It has separate checkpoints for handwritten (IAM), printed (SROIE) and scene text, and loads in a few lines with Hugging Face transformers.
Key differences
| Jina OCR | TrOCR | |
|---|---|---|
| Input scope | Whole pages and PDFs, output as Markdown | Single text-line images only; you need a separate line detector |
| Licence | CC-BY-NC-4.0; commercial use needs a license from Jina/Elastic | MIT (microsoft/unilm repository) |
| How you run it | Self-host or Jina's hosted API | Local only, through Hugging Face transformers |
| Output formats | Markdown, HTML, LaTeX | Plain text; no layout, table or page-level output |
| Model size | 3B MoE, about 570M active | Base 334M, large 558M |
| Benchmark | OmniDocBench v1.6 overall 91.14; olmOCR-Bench overall 83.4 (authors) | IAM handwriting CER 2.89 with TrOCR-large (authors) |
| Known failure | Hosted API returns HTTP 503 on cold start; retry after 30 to 60 seconds | Released checkpoints are English-focused; other languages need fine-tuning |
Sources
- jina-ocr-v1 model card
- Jina model page
- TrOCR paper (arXiv 2109.10282)
- trocr-large-handwritten model card
What each one is
Jina OCR
Compact page-to-Markdown model from Jina AI, built on the DeepSeek-OCR encoder.
Jina OCR detailsTrOCR
Transformer text recognition model. Reads single text lines, including handwriting.
TrOCR 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": ["microsoft/trocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Jina OCR alternativesTrOCR alternativesBest OCR models in 2026