Unlimited OCR vs Tesseract
Pick Unlimited OCR when you have a GPU and need to parse long reports or books in one request, with block types and bounding boxes for RAG or layout-aware extraction. Pick Tesseract when you need cheap CPU-only OCR of clean printed scans in many languages, or searchable PDFs offline. The main split is hardware and layout: Unlimited OCR needs a CUDA GPU and returns structure, while Tesseract runs anywhere and returns only lines and words.
Unlimited OCRbaidu/unlimited-ocr | Tesseracttesseract/tesseract | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | Free |
| Per 1,000 pages | $2.00 | $0.00 |
| Typical speed | 1.2 s / page | 0.6 s / page |
| Best for | Books and very long documents | Clean printed text, free |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | 100+ languages |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| 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 Unlimited OCR if
- It parses 40+ pages in a single forward pass with roughly constant memory, so long reports and books go through in one request.
- It emits Markdown with block types and bounding boxes, which suits layout-aware extraction and bulk PDF to text for RAG.
- Baidu reports 77.1 on MarathonDocBench for Unlimited OCR 2, against 33.7 for the next model.
Choose Tesseract if
- It runs on CPU only with no model server, so it fits small machines and offline use.
- It recognizes more than 100 languages out of the box and can be trained for more.
- It outputs plain text, hOCR, searchable PDF, TSV, ALTO and PAGE XML, and it is the engine behind OCRmyPDF.
Key differences
| Unlimited OCR | Tesseract | |
|---|---|---|
| Licence | MIT | Apache-2.0 |
| How you run it | Self-host with vLLM, SGLang or transformers; also hosted in Baidu Cloud's OCR product | Local CLI or C++ library with an LSTM recognizer |
| Hardware | NVIDIA GPU with CUDA 12.9 or 13.0; version 2 under SGLang runs out of memory on 12 GB GPUs | CPU only, no GPU needed |
| Price | Free self-hosted, you pay for compute | Free self-hosted, you pay for compute |
| Output formats | Markdown, plain text, bounding boxes | Plain text, hOCR, searchable PDF, TSV, ALTO, PAGE |
| Layout and tables | Block types and bounding boxes alongside text; formatting score on ParseBench is only 0.97, so bold and italic are mostly lost | No table structure, formula or reading-order output; lines and words only |
| Known failure | Text repetition and looping (issue #55), hallucinated text (issue #58), bbox coordinate loops on GB10 GPUs under SGLang (issue #114) | Accuracy drops sharply without binarization, deskew and roughly 300 DPI input; confuses digit 1 with lowercase l (issue #2753) |
Sources
- Hugging Face model card
- GitHub repo
- MarkTechPost on Unlimited OCR 2
- Tesseract GitHub README
- Improving quality (tessdoc)
- Issue #2753 1 vs l
What each one is
Unlimited OCR
Open-source model from Baidu that reads hundreds of pages in one pass with flat memory use.
Unlimited 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": "baidu/unlimited-ocr","fallbacks": ["tesseract/tesseract"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Unlimited OCR alternativesTesseract alternativesBest OCR models in 2026