SmolDocling vs Nougat
Pick Nougat for English scientific papers where you need math and tables as LaTeX, as long as the work is non-commercial. Pick SmolDocling for general English page conversion with a permissive licence, but treat it as a preview and plan to move to Granite-Docling. The main split is licence and scope: Nougat weights are CC-BY-NC and built for arXiv-style papers, while SmolDocling is Apache-2.0 and covers text, tables, code, formulas and charts.
SmolDoclinghuggingface/smoldocling | Nougatmeta/nougat | |
|---|---|---|
| Type | Open source | Open source |
| Price | 1 credit / page | 2 credits / page |
| Per 1,000 pages | $1.00 | $2.00 |
| Typical speed | 1.0 s / page | 3.0 s / page |
| Best for | Very cheap page conversion | arXiv-style papers |
| Inputs | PDF, images | PDF, images |
| Languages | Mainly English | 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 SmolDocling if
- The licence is Apache-2.0 per the model summary, so there is no non-commercial restriction on the weights.
- It is 256M parameters and runs at about 0.35 seconds per page on an A100 with vLLM, which suits modest hardware.
- One model outputs DocTags, Markdown, HTML, JSON, LaTeX and bounding boxes, and handles text, tables, code, formulas and charts with location tags.
Choose Nougat if
- It is trained on arXiv and PubMed Central papers, so it fits English scientific papers and math-heavy PDFs.
- It writes equations and tables as LaTeX inside Mathpix-compatible .mmd Markdown.
- It ships small and base checkpoints with a simple CLI and API, and the code is MIT licensed.
Key differences
| SmolDocling | Nougat | |
|---|---|---|
| Licence | Apache-2.0 per model summary (metadata tag says CDLA-Permissive-2.0) | Code MIT; weights CC-BY-NC, so no commercial use |
| Pricing | Free, self-hosted (you pay for compute) | Free for non-commercial use, self-hosted (you pay for compute) |
| Output formats | DocTags, Markdown, HTML, JSON, LaTeX, bounding boxes | Markdown (Mathpix-style .mmd) and LaTeX |
| Target documents | General page conversion: text, tables (OTSL), code, formulas, charts | Scientific PDF pages from arXiv and PubMed Central style papers |
| Languages | English only | Works best on English; Chinese, Russian and Japanese do not work |
| Benchmarks | Table TEDS structure on FinTabNet 0.82, full-page OCR F1 0.80, OCRBench 338 | Not published |
| Known failure | Endless token repetition and missing closing tags; Docling issue #1515 reports looping output that makes generation very slow | Repetition loops on 1.5% of test pages, more on out-of-domain documents; repetition detection can drop pages as MISSING_PAGE |
Sources
- SmolDocling model card
- Granite-Docling comparison table
- Docling issue #1515
- Nougat GitHub README
- Nougat paper
- Issue #11 repetitions
What each one is
SmolDocling
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": "huggingface/smoldocling","fallbacks": ["meta/nougat"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
SmolDocling alternativesNougat alternativesBest OCR models in 2026