OCRRouter

Chandra vs Google Cloud Vision OCR

Pick Chandra when you need structured output from scanned documents with tables, forms or math, especially in several languages, and you can run a GPU or pay the hosted API. Pick Google Cloud Vision OCR for high-volume plain text extraction, photos of signs and labels, or handwritten notes, where cost per page and a managed API matter more than layout. The main split is structure: Chandra returns Markdown, HTML and JSON with layout, while Cloud Vision returns text and bounding boxes with no table or Markdown output.

Chandradatalab/chandraGoogle Cloud Vision OCRgoogle/cloud-vision
TypeOpen sourceOCR API
Price2 credits / page2 credits / page
Per 1,000 pages$2.00$2.00
Typical speed2.0 s / page1.0 s / page
Best forComplex layouts and handwritingPhotos and simple scans
InputsPDF, imagesPDF, images
Languages40+ languages100+ languages
Max pages per request5002,000
Tables✓ YesNo
Handwriting✓ Yes✓ Yes
Math✓ YesNo
Bounding boxes✓ Yes✓ Yes
ConfidenceNo✓ Yes
Forms✓ YesNo
Multilingual✓ Yes✓ 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 Chandra if

  • You want Markdown, HTML or JSON with per-block bounding boxes and labels, for example to feed a RAG pipeline.
  • You process multilingual scans: Chandra 2 averages 77.8% across 43 languages, against 67.6% for Gemini 2.5 Flash in the authors' table.
  • You want to self-host on one NVIDIA GPU through vLLM, or use the hosted Datalab API at $1, $3 or $7.50 per 1,000 pages.

Choose Google Cloud Vision OCR if

  • You need cheap plain text extraction at volume: $1.50 per 1,000 units, falling to $0.60 above 5M units, with the first 1,000 units per month free.
  • You read handwritten notes or photos of signs and labels, using DOCUMENT_TEXT_DETECTION and language hints.
  • You want a managed Google Cloud API with no GPU to run, and a page, block, paragraph, word and symbol hierarchy with bounding polygons.

Key differences

ChandraGoogle Cloud Vision OCR
How you run itSelf-host the open weights (vLLM or Hugging Face) or use the hosted Datalab APIHosted API only, over REST or client libraries on Google Cloud; no on-premise option
LicenceCode Apache-2.0; weights under modified OpenRAIL-M, free for research, personal use and companies under $2M funding or revenueProprietary
PriceSelf-hosted: free within licence terms, you pay for compute. Hosted: $1, $3 or $7.50 per 1,000 pages; $20/month free tierFirst 1,000 units per month free, then $1.50 per 1,000 up to 5M, $0.60 per 1,000 above; one unit per PDF page
Output formatsMarkdown, HTML, JSON and bounding boxesJSON, plain text and bounding boxes; no Markdown, so you rebuild reading order yourself
Tables and layoutLayout chunks with bounding boxes and labels per block; listed as good for tables, forms and mathNo table or layout structure; Google points to Document AI for forms and tables
BenchmarkChandra 2: 85.8 on olmOCR-bench, 77.8% average on a 43-language benchmark (authors' figures)Not published
Known failureToken-loop hallucinations reported on 5-6 of about 60 Russian/Kazakh scans; weak on Urdu (43.2%) and Sanskrit (51.1%); 66 seconds per page measured on an RTX 4090 with the Hugging Face pathPDF and TIFF need async processing through Cloud Storage, and billing must be enabled even for the free 1,000 units

Sources

What each one is

Chandra

Datalab's OCR model that keeps layout and handles forms, tables and handwriting.

Chandra details

Google Cloud Vision OCR

Google Cloud's image OCR with text and document text detection, word boxes and confidence.

Google Cloud Vision OCR 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": "datalab/chandra","fallbacks": ["google/cloud-vision"],"file_url": "https://example.com/document.pdf"}

Questions

More comparisons

Chandra alternativesGoogle Cloud Vision OCR alternativesBest OCR models in 2026