Amazon Textract vs Chandra
Pick Amazon Textract for English forms, invoices and large scanned PDF batches on AWS, where you want a managed API with word and line geometry and confidence scores. Pick Chandra for multilingual documents or PDF to Markdown for RAG, since it outputs Markdown, HTML and JSON and scored 77.8% across 43 languages. Language coverage is the main split: Textract text detection stops at six Latin-script languages.
Amazon Textractaws/textract | Chandradatalab/chandra | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 25 credits / page | 2 credits / page |
| Per 1,000 pages | $25.00 | $2.00 |
| Typical speed | 1.2 s / page | 2.0 s / page |
| Best for | Forms and key-value extraction | Complex layouts and handwriting |
| Inputs | PDF, images | PDF, images |
| Languages | 6 languages | 40+ languages |
| Max pages per request | 3,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | ✓ Yes | ✓ Yes |
| 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 Amazon Textract if
- Your documents are English forms and invoices, and Handwriting and Queries only work on English.
- You run on AWS and want a hosted API that handles PDFs up to 500 MB and 3,000 pages through the async API.
- You need basic text detection at $1.50 per 1,000 pages, with word and line geometry and confidence scores.
Choose Chandra if
- You have scanned documents in many languages, or non-Latin scripts that Textract text detection does not support.
- You need Markdown, HTML or JSON with layout chunks for a RAG pipeline, without post-processing.
- You want to self-host open weights on one NVIDIA GPU, or use the hosted Datalab API at $1 to $7.50 per 1,000 pages.
Key differences
| Amazon Textract | Chandra | |
|---|---|---|
| Deployment | Hosted AWS API only | Self-host with vLLM or Hugging Face, or the hosted Datalab API |
| Licence | Proprietary | Code Apache-2.0; weights modified OpenRAIL-M, free for research, personal use and companies under $2M funding or revenue, paid commercial licence above that |
| Price | First 1M pages/month in US West (Oregon): $0.0015/page text, $0.015 Tables, $0.05 Forms, $0.015 Queries | Self-hosted: free within licence terms, you pay for compute. Hosted: $1, $3 or $7.50 per 1,000 pages; free tier gives $20/month of usage |
| Output formats | JSON, plain text, bounding boxes | Markdown, HTML, JSON, bounding boxes |
| Languages | Text detection: English, French, German, Italian, Portuguese, Spanish. Handwriting and Queries: English only. Vertical Japanese and Chinese not supported | 77.8% average across 43 languages and 72.7% across 90 languages (authors' benchmark). Weak on Urdu (43.2%) and Sanskrit (51.1%) |
| Benchmark | 80.9% on RD-TableBench table accuracy (independent) | 85.8 on olmOCR-bench overall for Chandra 2 (authors' figure). Old Scans category only 51.1 |
| Known failure | Sync API takes one page; large files need the async API | Token-loop hallucinations reported on 5-6 of about 60 Russian and Kazakh scans, and missing pages. One reviewer measured 66 seconds per page on an RTX 4090 with the Hugging Face path |
Sources
- Textract pricing
- Textract hard limits
- Reducto RD-TableBench
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
What each one is
Amazon Textract
AWS OCR with form key-value pairs, tables and per-word confidence scores.
Amazon Textract detailsChandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra 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": "aws/textract","fallbacks": ["datalab/chandra"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Amazon Textract alternativesChandra alternativesBest OCR models in 2026