OCRRouter

Mistral OCR vs GOT-OCR 2.0

Pick Mistral OCR for high-volume PDF to Markdown and batch scanning of business documents through a hosted API at $2 per 1,000 pages. Pick GOT-OCR 2.0 when you need to self-host and read formulas, sheet music, charts or chemical structures, or run region-specific OCR on English and Chinese documents. The main split is hosted API with per-page pricing versus a free 580M-parameter model that you run on your own GPU.

Mistral OCRmistral/mistral-ocrGOT-OCR 2.0stepfun/got-ocr-2
TypeOCR APIOpen source
Price25 credits / page1 credit / page
Per 1,000 pages$25.00$1.00
Typical speed0.9 s / page1.5 s / page
Best forHigh-volume PDF to MarkdownFormatted text, formulas and charts
InputsPDF, imagesimages
Languages100+ languagesChinese, English
Max pages per request1,000100
Tables✓ Yes✓ Yes
HandwritingNoNo
Math✓ Yes✓ Yes
Bounding boxes✓ Yes✓ Yes
ConfidenceNoNo
FormsNoNo
Multilingual✓ YesNo
On OCRRouterAvailableAvailable

Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.

Which to choose

Choose Mistral OCR if

  • You want a hosted API with no GPU to manage, at $2 per 1,000 pages, or $1 per 1,000 through the Batch API.
  • You need tables rebuilt as HTML with rowspan and colspan, and Markdown, HTML or JSON output with extracted images.
  • You process large batches of scanned business documents or feed PDFs into a RAG pipeline.

Choose GOT-OCR 2.0 if

  • You need to self-host. The code is Apache-2.0 and the model is free, so you pay only for compute.
  • You read formulas, sheet music, charts or chemical structures, and want LaTeX, TikZ, SMILES or kern output.
  • You need region OCR guided by box coordinates or color, or multi-page OCR with batched inference in Hugging Face Transformers.

Key differences

Mistral OCRGOT-OCR 2.0
How you run itHosted API (current pinned version mistral-ocr-2512, OCR 3)Self-hosted; available in Hugging Face Transformers
LicenceProprietaryCode Apache-2.0; training data CC BY-NC 4.0
Price$2 per 1,000 pages; $3 with structured annotations; $1 per 1,000 via Batch APIFree; you pay for compute
Output formatsMarkdown, HTML, JSON, imagesPlain text, Markdown, LaTeX, TikZ, SMILES, kern
BenchmarkOmniDocBench v1.5 overall 78.83 (Mistral OCR 2503); v1.6 overall 85.66Not published
HardwareNone to manage; it runs on the APICUDA 11.8 and PyTorch 2.0.1 per README; CPU users pointed to a llama.cpp port
Known failuresColored-background pages returned as images with no OCR text (2025 release); hallucinated text on unreadable handwriting; changed checkbox states and misread characters on a dense financial tableOutput can loop characters midway through recognition; wrong text instead of an empty string on images with no text; format mode gives much worse results than plain OCR mode

Sources

What each one is

Mistral OCR

Fast document OCR that returns clean Markdown with tables and embedded images.

Mistral OCR details

GOT-OCR 2.0

Unified OCR model that reads plain text, tables, formulas, sheet music and charts.

GOT-OCR 2.0 details

Try 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": "mistral/mistral-ocr","fallbacks": ["stepfun/got-ocr-2"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Mistral OCR alternativesGOT-OCR 2.0 alternativesBest OCR models in 2026