Docling vs docTR
Pick Docling when you need to turn PDFs, Office files, HTML or images into structured Markdown or JSON, for example in a RAG pipeline. Pick docTR when you only need word-level OCR with coordinates, for example on invoices and receipts that feed your own extraction code. The main split is scope: Docling converts whole documents and leans on a pluggable OCR engine for scans, while docTR is a text detection and recognition library that returns words and lines only.
Doclingibm/docling | docTRmindee/doctr | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 2.2 s / page | 0.9 s / page |
| Best for | Structured export for RAG | Word-level OCR with positions |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | Latin scripts, CJK | Mainly Latin scripts |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | No |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | ✓ Yes |
| 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
- It reads PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails and XBRL through one API.
- Its lossless JSON export keeps layout, reading order and table structure, and it also writes Markdown, HTML, DocTags and plain text.
- It plugs into EasyOCR, Tesseract, RapidOCR and VLMs such as the 258M GraniteDocling, and it has LangChain and LlamaIndex integrations.
Choose docTR if
- It returns a nested page, block, line and word structure with coordinates as JSON, which suits custom extraction on invoices and receipts.
- You can choose the detection and recognition architectures, use straight or rotated boxes, and train your own models.
- It is Apache-2.0, ships ready Docker images, has a key information extraction predictor, and has a demo on Hugging Face Spaces.
Key differences
| Docling | docTR | |
|---|---|---|
| What it is | Document converter that outputs a structured DoclingDocument and plugs in OCR engines and VLMs | Two-stage text detection and recognition OCR library |
| Licence | MIT (individual models have their own licences) | Apache-2.0 |
| Input formats | PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, emails, XBRL | Images and documents for OCR, no other formats documented |
| Output | Markdown, HTML, JSON, DocTags, plain text | JSON, plain text, bounding boxes. No table, formula or Markdown output |
| Hardware | macOS, Linux and Windows on x86_64 and arm64. CPU works, GPU speeds up model stages | CPU or NVIDIA GPU. GPU Docker images need CUDA 12.2 or newer on the host |
| Scanned pages | Accuracy depends on the OCR engine you plug in | Cropping images leads to poorer results (issue #2007) |
| Known issues | Version 2.143.1 OCR regression drops words and numbers with full-page Tesseract (#5254). Reading order can merge neighbouring columns (#5132). Over 1,000 open issues | Split-crop merging misjoins repeated characters (#2229). ONNX export of MASTER truncates decoding steps (#2224) |
Sources
- Docling GitHub README
- Supported formats
- Issue #5254 OCR regression
- docTR GitHub README
- Issue #2229
- Issue #2007
What each one is
docTR
Open-source OCR library from Mindee with separate text detection and recognition models.
docTR 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": "ibm/docling","fallbacks": ["mindee/doctr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Docling alternativesdocTR alternativesBest OCR models in 2026