Nanonets-OCR vs SmolDocling
Pick Nanonets-OCR for printed business documents where you need signatures, watermarks, checkboxes or flowcharts turned into structured output, and accept that its olmOCR-bench overall score is 69.5. Pick SmolDocling only for cheap English experiments with a 256M model that gives bounding boxes. It is a preview release, so move to Granite-Docling for anything in production.
Nanonets-OCRnanonets/nanonets-ocr | SmolDoclinghuggingface/smoldocling | |
|---|---|---|
| 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 | 1.0 s / page |
| Best for | Markdown with semantic tags | Very cheap page conversion |
| Inputs | PDF, images | PDF, images |
| Languages | Multilingual | Mainly English |
| 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
- It tags signatures, watermarks and image descriptions, and converts checkboxes to unicode symbols.
- It extracts flowcharts and org charts as Mermaid code and tells inline LaTeX from display LaTeX.
- The authors report DocVQA 89.43 and ChartQA 78.56 for the OCR2-3B model.
Choose SmolDocling if
- At 256M parameters it runs on modest hardware, and the authors report about 0.35 seconds per page on an A100 with vLLM.
- It returns DocTags with location tags and bounding boxes, plus Markdown, HTML and JSON.
- It carries a stated Apache-2.0 licence, while the Nanonets-OCR2-3B model card states none.
Key differences
| Nanonets-OCR | SmolDocling | |
|---|---|---|
| Model size | 3B (Hugging Face lists about 4B params) | 256M |
| Licence | Not stated on the Nanonets-OCR2-3B model card; base model is Qwen2.5-VL-3B | Apache-2.0 per model summary (metadata tag says CDLA-Permissive-2.0) |
| Output formats | Markdown, HTML, LaTeX, Mermaid | DocTags, Markdown, HTML, JSON, LaTeX, bounding boxes |
| Reported benchmarks | olmOCR-bench overall 69.5, DocVQA 89.43, ChartQA 78.56 (OCR2-3B, by the authors) | Table TEDS 0.82 on FinTabNet 150dpi, full-page OCR F1 0.80, OCRBench 338 (by the authors) |
| Handwriting | Not trained on handwriting and can hallucinate. One test saw generation loops with CER up to 2100% and 69.1s per image on an L4. | Not published |
| Languages | Not published | English only |
| Known failure | Fixed prompt only. The team expects hallucination with other prompts. | Endless token repetition and missing closing tags. Output can loop on some images and become very slow. |
Sources
- Nanonets-OCR2-3B model card
- Nanonets-OCR-s announcement
- HF discussion fixed prompt
- SmolDocling model card
- Docling issue #1515
- Granite-Docling comparison table
What each one is
Nanonets-OCR
Open image-to-Markdown model that tags tables, equations, signatures, watermarks and checkboxes.
Nanonets-OCR detailsSmolDocling
Tiny 256M vision model from Hugging Face and IBM that outputs DocTags. Predecessor of Granite Docling.
SmolDocling 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": ["huggingface/smoldocling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Nanonets-OCR alternativesSmolDocling alternativesBest OCR models in 2026