Docling vs Nougat
Pick Docling for general document ingestion: PDFs, Office files, HTML and images into structured JSON, under an MIT licence that allows commercial use. Pick Nougat only for English scientific papers where you need LaTeX math and tables, and only for non-commercial work, because its weights are CC-BY-NC.
Doclingibm/docling | Nougatmeta/nougat | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.2 s / page | 3.0 s / page |
| Best for | Structured export for RAG | arXiv-style papers |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | Latin scripts, CJK | Mainly English |
| Max pages per request | 500 | 200 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | No | No |
| Multilingual | No | 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 Docling if
- You need one API for PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL.
- You are building a commercial product, since the code is MIT (individual models carry their own licences).
- You want lossless JSON that keeps layout, reading order and table structure, or you use LangChain or LlamaIndex.
Choose Nougat if
- Your inputs are English arXiv-style papers and you need equations and tables as LaTeX in Mathpix-compatible Markdown.
- Your work is non-commercial, so the CC-BY-NC weights are acceptable.
- You want a small or base checkpoint with a simple CLI and API, trained on arXiv and PubMed Central documents.
Key differences
| Docling | Nougat | |
|---|---|---|
| Licence | MIT (individual models have their own licences) | Code MIT; weights CC-BY-NC, so no commercial use |
| Input formats | PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails, XBRL | Scientific PDF pages |
| Output formats | Markdown, HTML, JSON, DocTags, plain text | Markdown (.mmd) and LaTeX |
| How OCR works | Pluggable engines (EasyOCR, Tesseract, RapidOCR) and VLMs such as the 258M GraniteDocling; scan accuracy depends on the engine you plug in | Single Visual Transformer model, trained on arXiv and PubMed Central |
| Languages | Not published | Best on English; README says Chinese, Russian and Japanese do not work |
| Pricing | Free, self-hosted (you pay for compute) | Free for non-commercial use, self-hosted (you pay for compute) |
| Known failure | Tesseract full-page OCR in 2.143.1 drops words and numbers (issue #5254); reading order can merge neighbouring columns (issue #5132) | Repetition loops on 1.5% of test pages in the paper, more on out-of-domain documents; looping pages can be dropped as MISSING_PAGE (issue #11) |
Sources
- Docling GitHub README
- Supported formats
- Issue #5254 OCR regression
- Nougat GitHub README
- Nougat paper
- Issue #11 repetitions
What each one is
Try 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": ["meta/nougat"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling vs DonutDocling vs MonkeyOCRTesseract vs NougatMinerU vs NougatLlamaParse vs DoclingPaddleOCR-VL vs DoclingMinerU vs DoclingDocling vs Tesseract
Docling alternativesNougat alternativesBest OCR models in 2026