Mathpix Convert vs Unlimited OCR
Pick Mathpix Convert for math-heavy papers and textbooks where you need LaTeX or DOCX out of a PDF, including handwritten equations. Pick Unlimited OCR for long reports and books that you want parsed to Markdown with bounding boxes, on your own NVIDIA GPU under an MIT licence. The main split is a paid hosted STEM API against a free self-hosted general document parser.
Mathpix Convertmathpix/convert | Unlimited OCRbaidu/unlimited-ocr | |
|---|---|---|
| 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 | 1.2 s / page |
| Best for | Math and scientific papers | Books and very long documents |
| Inputs | PDF, images | PDF, images |
| Languages | 30+ languages | 30+ languages |
| Max pages per request | 500 | 1,000 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| 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 content, turning math, tables and chemistry into LaTeX and Markdown.
- It exports whole PDFs directly to DOCX and LaTeX, which Unlimited OCR does not output.
- It handles handwritten equations and needs no GPU, because it runs as a hosted API.
Choose Unlimited OCR if
- The MIT licence allows commercial use with no copyleft, and the model is free to self-host.
- It parses 40+ pages in a single forward pass with roughly constant memory, so long reports and books go in one request.
- It returns block types and bounding boxes alongside text, which suits layout-aware extraction and bulk PDF to text for RAG.
Key differences
| Mathpix Convert | Unlimited OCR | |
|---|---|---|
| Licence | Proprietary | MIT |
| How you run it | Hosted API only | Self-host with vLLM, SGLang or transformers, or use the hosted endpoint in Baidu Cloud's OCR product |
| Price | PDF $0.005 per page up to 1M pages, then $0.0035; images $0.002 each; one-time $19.99 API key fee | Free to self-host, you pay for compute; hosted price Not published |
| Output formats | LaTeX, Markdown, DOCX, HTML, JSON | Markdown, plain text, bounding boxes |
| Benchmarks | No published accuracy benchmark on general OCR leaderboards | ParseBench mean 46.17 (independent); Unlimited OCR 2 scores 77.1 on MarathonDocBench and 84.6 on MPDocBench (author-reported) |
| Hardware | None on your side | NVIDIA GPU with CUDA 12.9 or 13.0; Unlimited OCR 2 under SGLang runs out of memory on 12 GB GPUs |
| Known failure | Images with more than 12 text rows may be billed at the PDF page rate | Text repetition and looping, reported hallucinated text, and bold or italic styling mostly lost |
Sources
- Mathpix billing docs
- Mathpix OCR API
- Hugging Face model card
- MarkTechPost on Unlimited OCR 2
- Issue #116 OOM on 12 GB
What each one is
Mathpix Convert
Specialist for math, science papers and handwritten formulas. Outputs LaTeX.
Mathpix Convert detailsUnlimited OCR
Open-source model from Baidu that reads hundreds of pages in one pass with flat memory use.
Unlimited OCR 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": ["baidu/unlimited-ocr"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Mathpix Convert alternativesUnlimited OCR alternativesBest OCR models in 2026