OCRRouter

Amazon Textract vs Surya

Pick Amazon Textract for English forms, invoices and large scanned PDF batches if you already run on AWS and want a managed API with per-page pricing. Pick Surya for local or self-hosted OCR with layout, reading order and bounding boxes across 90+ languages, when you can run your own inference server. The main split is language coverage and deployment: Textract is a hosted API limited to six Latin-script languages, while Surya is open source and self-hosted.

Amazon Textractaws/textractSuryadatalab/surya
TypeOCR APIOpen source
Price25 credits / page2 credits / page
Per 1,000 pages$25.00$2.00
Typical speed1.2 s / page1.5 s / page
Best forForms and key-value extractionLayout and line-level OCR
InputsPDF, imagesPDF, images
Languages6 languages90+ languages
Max pages per request3,000500
Tables✓ Yes✓ Yes
Handwriting✓ YesNo
MathNoNo
Bounding boxes✓ Yes✓ Yes
Confidence✓ YesNo
Forms✓ YesNo
MultilingualNo✓ Yes
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 Amazon Textract if

  • Text detection costs $0.0015 per page ($1.50 per 1,000 pages) in US West (Oregon) for the first 1M pages a month, with no servers to run.
  • The async API takes PDFs up to 500 MB and 3,000 pages, which suits large scanned batches.
  • It extracts forms ($0.05/page), tables ($0.015/page) and Queries ($0.015/page), and returns word and line geometry with confidence scores.

Choose Surya if

  • It covers 90+ languages, while Textract text detection supports only English, French, German, Italian, Portuguese and Spanish.
  • Code is Apache-2.0 and the software is free to self-host, so you pay only for compute, as long as your funding or revenue stays under $5M.
  • One 689M-parameter model returns layout labels, reading order, OCR, HTML tables and KaTeX math with per-block bounding boxes, and it can run on CPU or Apple Silicon through llama.cpp.

Key differences

Amazon TextractSurya
DeploymentHosted AWS APISelf-host; needs a separate vLLM or llama.cpp inference server
LicenceProprietaryCode Apache-2.0; weights under modified AI Pubs Open Rail-M, free below $5M funding/revenue
PriceText detection $0.0015/page; Tables $0.015; Forms $0.05; Queries $0.015 (first 1M pages/month, US West Oregon)Free software; you pay for compute; commercial licence needed above $5M revenue or funding
LanguagesText detection in 6 Latin-script languages; handwriting and Queries English only; vertical Japanese and Chinese not supported90+ languages, but only 21 of 91 reach a 90% pass rate and multilingual tables score 58.1
Tables80.9% on RD-TableBench (independent)Surya 2.5 table output lacks row/column geometry; simple table mode errors and needs the older Surya 2 checkpoint
Output formatsJSON, plain text, bounding boxesJSON, HTML, LaTeX, bounding boxes, plain text
Limits and speedSync API takes one page; async handles up to 3,000 pages4.52 pages/s at 192 DPI on one H100 with vLLM; 0.198 pages/s on an M4 Max with llama.cpp at 96 DPI

Sources

What each one is

Amazon Textract

AWS OCR with form key-value pairs, tables and per-word confidence scores.

Amazon Textract details

Surya

Open-source OCR toolkit with line detection, layout analysis, reading order and tables.

Surya 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": "aws/textract","fallbacks": ["datalab/surya"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Amazon Textract alternativesSurya alternativesBest OCR models in 2026