Mathpix Convert vs SmolDocling
Pick Mathpix Convert for math-heavy papers and textbooks, where you need LaTeX, DOCX or Markdown from whole PDFs, including handwritten equations. Pick SmolDocling only for English experiments on your own hardware, since it is a free 256M preview model that the authors say is superseded by Granite-Docling. The main split is a paid hosted STEM API against a free self-hosted preview model.
Mathpix Convertmathpix/convert | SmolDoclinghuggingface/smoldocling | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 7 credits / page | 1 credit / page |
| Per 1,000 pages | $7.00 | $1.00 |
| Typical speed | 1.5 s / page | 1.0 s / page |
| Best for | Math and scientific papers | Very cheap page conversion |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | Mainly English |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | 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 Mathpix Convert if
- You convert math, tables and chemistry in papers and textbooks to LaTeX or Markdown.
- You need DOCX or LaTeX export directly from whole PDFs, or handwritten equations read.
- You want a hosted API with no GPU to run, at $0.005 per PDF page and $0.002 per image.
Choose SmolDocling if
- You want to run a tiny 256M-parameter model on modest hardware and pay only for compute.
- You need Apache-2.0 licensed weights (per the model summary) and output with bounding boxes and DocTags for Docling.
- You are testing a small document VLM on English pages, at about 0.35 seconds per page on an A100 with vLLM.
Key differences
| Mathpix Convert | SmolDocling | |
|---|---|---|
| How you run it | Hosted API | Self-hosted, 256M parameters |
| Licence | Proprietary | Apache-2.0 per the model summary; the metadata tag says CDLA-Permissive-2.0 |
| Price | PDF $0.005 per page (0-1M), $0.0035 above 1M; images $0.002 each; one-time $19.99 API key fee | Free; you pay for compute |
| Output formats | LaTeX, Markdown, DOCX, HTML, JSON | DocTags, Markdown, HTML, JSON, LaTeX, bounding boxes |
| Languages | Not published | English only |
| Benchmarks | No published accuracy benchmark on general OCR leaderboards | Table TEDS 0.82 (FinTabNet 150dpi), full-page OCR F1 0.80, OCRBench 338, all reported by the authors |
| Known failure | Images with more than 12 text rows may be billed at the PDF page rate | Endless token repetition, missing closing tags, and output loops on some images (Docling issue #1515) |
Sources
- Mathpix billing docs
- Mathpix OCR API
- SmolDocling model card
- Docling issue #1515
- Granite-Docling comparison table
What each one is
Mathpix Convert
Specialist for math, science papers and handwritten formulas. Outputs LaTeX.
Mathpix Convert 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": "mathpix/convert","fallbacks": ["huggingface/smoldocling"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Mathpix Convert alternativesSmolDocling alternativesBest OCR models in 2026