dots.ocr vs Granite Docling
Pick dots.ocr for multilingual PDFs with mixed tables and text, where you have a CUDA GPU and want layout boxes, HTML tables and LaTeX formulas from one model. Pick Granite Docling for English PDFs with tables, code and equations inside a Docling pipeline, or when you need a 258M model that runs cheaply on CPU or a small GPU. The main split is size and language: 3B and multilingual versus 258M and English-first.
dots.ocrrednote/dots-ocr | Granite Doclingibm/granite-docling | |
|---|---|---|
| Type | Open source | Open source |
| Price | 2 credits / page | 1 credit / page |
| Per 1,000 pages | $2.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Multilingual document parsing | Small-model document conversion |
| Inputs | PDF, images | PDF, images |
| Languages | 100+ languages | English, experimental CJK and Arabic |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| 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 dots.ocr if
- It handles multilingual documents, while Granite Docling marks Japanese, Arabic and Chinese as experimental.
- dots.mocr (v1.5) scores 83.9 on olmOCR-bench, up from 79.1 for the original dots.ocr.
- The dots.mocr-svg variant also converts charts and diagrams to SVG.
Choose Granite Docling if
- At 258M parameters it suits low-cost CPU or small GPU conversion, against 3B for dots.ocr.
- It reports table structure TEDS of 0.97 on FinTabNet, code F1 of 0.988 and equation F1 of 0.968.
- It is Apache-2.0 licensed and built to feed the Docling library through DocTags.
Key differences
| dots.ocr | Granite Docling | |
|---|---|---|
| Size | 3B (1.7B LLM backbone) | 258M |
| Licence | MIT (repository also includes a separate dots.ocr license agreement file) | Apache-2.0 |
| How you run it | Self-hosted with vLLM or Hugging Face Transformers; single CUDA GPU with vLLM is the documented path | Self-hosted, typically inside Docling |
| Output formats | JSON, Markdown, HTML, LaTeX, bounding boxes, SVG | DocTags, Markdown, HTML, LaTeX, bounding boxes |
| Languages | Multilingual | English primary; Japanese, Arabic and Chinese experimental |
| Known failure | Runs of ellipses or underscores can make output repeat endlessly; complex tables and formulas remain difficult; a table above 1000 grid cells raises an error and drops the page (Docling issue #5146) | Repetitive loops on page headings and invoices; may output only exclamation marks on GPUs without bfloat16 such as T4 unless run in float32 |
| Price | Free, self-hosted (you pay for compute) | Free, self-hosted (you pay for compute) |
Sources
- dots.ocr GitHub README
- dots.ocr blog / limitations
- Granite-Docling model card
- HF discussion #24 repetition
- IBM announcement
What each one is
dots.ocr
Single vision model from rednote that does layout detection and content recognition together.
dots.ocr detailsGranite Docling
Compact IBM vision model that converts pages to DocTags for the Docling pipeline.
Granite Docling 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": "rednote/dots-ocr","fallbacks": ["ibm/granite-docling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
dots.ocr alternativesGranite Docling alternativesBest OCR models in 2026