Sample pricing card · numbers are estimates · not live

Chandra OCR 2

by Datalab · open weights

Turns scanned pages, PDFs and photos into clean Markdown, HTML or JSON. It keeps tables, forms, handwriting and reading order, which makes it a good fit for invoices, KYC files and handwritten records.

OCR · document layout Images and PDF pages in Hosted in India Separate instance No training on your data
Size4B parameters
RuntimevLLM, OpenAI-compatible
BillingFixed monthly plan
Accuracy on olmOCR-Bench ↗Ranked #1 of open OCR models
Chandra OCR 285.9%
dots.ocr 1.583.9%
olmOCR 282.4%
PaddleOCR-VL80.0%
DeepSeek-OCR75.4%

1. How many pages a month?

A page is one image or one PDF page. Rough guide: 1 invoice is 1–2 pages, 1 KYC file is 4–8 pages.

2. How fast do you need results?

3. When do you need it?

4. Extra usage

₹100 per 1,000 pages above your plan

5. Start your plan

Key terms

Your data
Processed in India on infrastructure dedicated to your plan. Documents and outputs are not stored after processing and are never used to train models.
Billing
A fixed monthly price plus 18% GST. Extra pages are billed only if you allow them, and never above your monthly limit.
Over your plan
With extra usage off, requests return HTTP 429 until the next month.
Price changes
We tell you about any new price before the month it applies to.
Cancellation
Pause or cancel at any time. Cancellation takes effect at the end of the current month.
Model licence
Your organisation is the licensee of the model. aram keeps a record of your confirmation.