Google Document AI vs Chandra
Pick Google Document AI for bulk OCR on Google Cloud, where $1.50 per 1,000 pages and 500-page batch requests matter more than table fidelity. Pick Chandra for multilingual scans and PDF to Markdown for RAG, especially if you want to self-host. The main split is a managed Google API with a 15-page online limit and weak table accuracy versus an open-weights model you can run on one GPU.
Google Document AIgoogle/document-ai | Chandradatalab/chandra | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 10 credits / page | 2 credits / page |
| Per 1,000 pages | $10.00 | $2.00 |
| Typical speed | 1.4 s / page | 2.0 s / page |
| Best for | Multilingual scans | Complex layouts and handwriting |
| Inputs | PDF, images | PDF, images |
| Languages | 200+ languages | 40+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | No | ✓ Yes |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | ✓ Yes | No |
| Forms | ✓ Yes | ✓ Yes |
| Multilingual | ✓ Yes | ✓ 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 Google Document AI if
- You already run on Google Cloud and want bulk OCR at $1.50 per 1,000 pages, dropping to $0.60 above 5M pages a month.
- You need RAG chunking from a managed service: Layout Parser includes it at $10 per 1,000 pages.
- You process large jobs in batch: batch requests take up to 500 pages.
Choose Chandra if
- You work with multilingual scans: Chandra 2 averages 77.8% across 43 languages in the authors' benchmark.
- You want to self-host on a single NVIDIA GPU via vLLM, with open weights and Apache-2.0 code.
- You need Markdown, HTML or JSON with bounding boxes per block, for documents with tables, forms and math.
Key differences
| Google Document AI | Chandra | |
|---|---|---|
| How you run it | Hosted API only | Self-host the open weights (vLLM or Hugging Face) or use the hosted Datalab API |
| Licence | Proprietary | Code Apache-2.0; weights modified OpenRAIL-M, free for research, personal use and companies under $2M funding/revenue |
| Price | Enterprise Document OCR $1.50 per 1,000 pages ($0.60 above 5M/month, first 1,000 free); Layout Parser $10; Form Parser $30 | Self-hosted: free within license terms, you pay for compute. Datalab API: $1, $3 or $7.50 per 1,000 pages by mode |
| Output formats | JSON, plain text, bounding boxes, chunks | Markdown, HTML, JSON, bounding boxes |
| Request limits | Online requests up to 15 pages (30 in imageless mode); batch up to 500 pages; Form Parser batch up to 100 pages | Not published |
| Table accuracy | 64.6% on RD-TableBench (independent), behind Azure and Textract | Not published |
| Known failure | Weak on complex tables; not suited to long PDFs in one synchronous call | Token-loop hallucinations reported on some Russian/Kazakh scans; 43.2% on Urdu, 51.1% on Sanskrit; Old Scans only 51.1 on olmOCR-bench |
Sources
- Document AI pricing
- Document AI limits
- Reducto RD-TableBench
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
What each one is
Google Document AI
Google Cloud OCR with layout parsing and specialised processors.
Google Document AI 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": "google/document-ai","fallbacks": ["datalab/chandra"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Google Document AI alternativesChandra alternativesBest OCR models in 2026