Mathpix Convert vs Surya
Pick Mathpix Convert for math-heavy papers and textbooks where you need LaTeX or DOCX from whole PDFs, including handwritten equations. Pick Surya for local, self-hosted OCR with layout, reading order and bounding boxes across 90+ languages. The main split is a hosted per-page API against a free model you run yourself.
Mathpix Convertmathpix/convert | Suryadatalab/surya | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 7 credits / page | 2 credits / page |
| Per 1,000 pages | $7.00 | $2.00 |
| Typical speed | 1.5 s / page | 1.5 s / page |
| Best for | Math and scientific papers | Layout and line-level OCR |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | 90+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | No | ✓ Yes |
| 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
- It is built for STEM and turns math, tables and chemistry into LaTeX and Markdown.
- It exports DOCX and LaTeX directly from whole PDFs.
- It is a hosted API, so you run no GPU or inference server, and it is listed as good for handwritten equations.
Choose Surya if
- It is free to self-host (Apache-2.0 code), with no per-page fee, until you pass $5M in funding or revenue and need a commercial license.
- It runs on your own hardware: an NVIDIA GPU with vLLM, or CPU and Apple Silicon with llama.cpp, at about 4.3-5.9 GB peak memory on an M4 Max.
- One 689M-parameter model returns layout labels, reading order, OCR, HTML tables and KaTeX math with per-block bounding boxes, and it scores 83.7% on olmOCR-bench (authors' figure).
Key differences
| Mathpix Convert | Surya | |
|---|---|---|
| How you run it | Hosted API | Self-host; needs a separate vLLM or llama.cpp inference server |
| Licence | Proprietary | Code Apache-2.0; weights under a modified AI Pubs Open Rail-M licence, free for research, personal use and companies under $5M funding or revenue |
| Price | PDF $0.005 per page (0-1M), $0.0035 above 1M; images $0.002 each; one-time $19.99 API key activation fee | Free; you pay for compute; commercial license needed above $5M |
| Output formats | LaTeX, Markdown, DOCX, HTML, JSON | JSON, HTML, LaTeX, bounding boxes, plain text |
| Benchmark | No published accuracy benchmark on general OCR leaderboards | 83.7% on olmOCR-bench and 85.9 on OmniParseBench (Surya 2.5, authors' figures) |
| Handwriting | Listed as good for handwritten equations | Listed as not for handwriting; the README suggests the hosted Datalab API for handwriting, complex tables and math |
| Known failure | Images with more than 12 text rows may be billed at the PDF page rate; billed monthly in arrears | Surya 2.5 table output lacks row/column geometry; only 21 of 91 languages reach 90% pass rate, and multilingual tables score 58.1 |
Sources
What each one is
Mathpix Convert
Specialist for math, science papers and handwritten formulas. Outputs LaTeX.
Mathpix Convert detailsSurya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "mathpix/convert","fallbacks": ["datalab/surya"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Mathpix Convert vs dots.ocrMathpix Convert vs ChandraReducto Parse vs SuryaLlamaParse vs SuryaMathpix Convert vs DeepSeek-VL2Surya vs NougatMathpix Convert vs Adobe Acrobat OCRQianfan OCR vs Surya
Mathpix Convert alternativesSurya alternativesBest OCR models in 2026