Nanonets-OCR vs Granite Docling
Pick Granite Docling for English PDFs with tables, code and equations, especially if you already run Docling or want a 258M model that is Apache-2.0 licensed and cheap to host. Pick Nanonets-OCR for printed business documents that carry signatures, watermarks and checkboxes, or when you want flowcharts as Mermaid. The main split is scope: Granite Docling is a small, structure-focused converter with a clear licence, while Nanonets-OCR is a larger model with more document-element tags but no stated licence.
Nanonets-OCRnanonets/nanonets-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 | 1.8 s / page | 1.5 s / page |
| Best for | Markdown with semantic tags | Small-model document conversion |
| Inputs | PDF, images | PDF, images |
| Languages | Multilingual | English, experimental CJK and Arabic |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | ✓ Yes | 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 Nanonets-OCR if
- You need signatures, watermarks and image descriptions tagged, and checkboxes converted to unicode symbols.
- You want flowcharts and org charts extracted as Mermaid code.
- You convert printed PDFs to Markdown for LLMs and want inline and display LaTeX equations told apart.
Choose Granite Docling if
- You need a licence you can read at a glance: Apache-2.0, versus no licence stated on the Nanonets-OCR2-3B model card.
- You process English PDFs with tables, code and equations: table structure TEDS is 0.97 on FinTabNet, code F1 is 0.988 and equation F1 is 0.968.
- You want to run on CPU or a small GPU, since the model has only 258M parameters, and you already use the Docling library.
Key differences
| Nanonets-OCR | Granite Docling | |
|---|---|---|
| Licence | Not stated on the Nanonets-OCR2-3B model card; base model is Qwen2.5-VL-3B | Apache-2.0 |
| Size | 3B (Hugging Face lists about 4B) | 258M |
| Output formats | Markdown, HTML, LaTeX, Mermaid | DocTags, Markdown, HTML, LaTeX, bounding boxes |
| Benchmarks (published by the authors) | olmOCR-bench overall 69.5; DocVQA 89.43; ChartQA 78.56 | Table TEDS 0.97 (FinTabNet); full-page OCR F1 0.84; OCRBench 500 |
| Handwriting and language | Not trained on handwriting and can hallucinate; one handwriting test saw generation loops with CER up to 2100% | English primary; Japanese, Arabic and Chinese are marked experimental |
| Prompting | Trained on a fixed prompt; hallucination with other prompts is expected | Outputs DocTags for the Docling library, typically run inside Docling |
| Known failure | Loses to Gemini 2.5 Flash in 52.43% of head-to-head Markdown comparisons, wins 39.98% | Repetitive loops on page headings and invoices; may output only exclamation marks on GPUs without bfloat16 such as T4 unless run in float32 |
Sources
- Nanonets-OCR2-3B model card
- Nanonets-OCR-s announcement
- HF discussion fixed prompt
- Granite-Docling model card
- HF discussion #24 repetition
What each one is
Nanonets-OCR
Open image-to-Markdown model that tags tables, equations, signatures, watermarks and checkboxes.
Nanonets-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": "nanonets/nanonets-ocr","fallbacks": ["ibm/granite-docling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Nanonets-OCR alternativesGranite Docling alternativesBest OCR models in 2026