Pay for pages. Nothing else.
Move the slider: the plan and per-page price adjust to your volume.
Try it free: your first 100 pages
An API key right after sign-up. Pick a plan later, once you know your volume.
We count the amount of work, not the number of sheets: a scanned or photographed page counts as two, an A1 drawing as eight, A2 as four, A3 as two. Translation and bundle splitting are included.
That is 1,800 pages per month.
- Plan
- Starter
- Total per month · VAT not applicable
- $0
- Price per page
- $0.000
Starter
your plan
$59/mo
early-access pricing
2,500 pages per month
For a product with early traffic
Pro
$399/mo
early-access pricing
20,000 pages per month
For production document volume
Business
custom
volume to match your load
On-premise, self-hosted LLM, SLA
What counts as a page
A page is a unit of work, not a physical sheet: max(text / 2,000 characters, sheet area / A4 format). An ordinary A4 text document is counted by characters. A drawing or a large-format sheet is counted by area: an A1 sheet weighs as much as eight A4 pages, A2 as four, A3 as two. A scan, a photo or a PDF without a text layer goes through OCR and counts as two pages. The result counts too: every 2,000 characters of the answer is a page, and you never get fewer pages than that, so a wide item schema costs what you asked to extract. Translation and splitting bundles into documents are already included.
page = max(text ÷ 2000, area ÷ А4, result ÷ 2000)
А1 = 8 pages
scan = page × 2What every page includes
- An in-house OCR loop, accurate on Cyrillic
- NDA boundary: PII scrubbing before any external LLM, fail_closed mode
- A model fallback cascade: the task survives provider outages
- Stateless: the document is not stored after processing
Business: the whole loop stays yours
On-premise deployment, self-hosted LLM, volumes and SLA matched to your case. Data never leaves your infrastructure at all.
Pricing questions
How a page is billed, what's included, and what happens with non-standard documents.
A page isn't a physical sheet — it's a unit of work: max(text / 2,000 characters, sheet area / A4 format). For ordinary A4 text documents that's a character count. For drawings and large-format sheets, area matters more: an A1 sheet weighs as much as eight A4 pages, A2 as four, A3 as two. A scan, a photo or a PDF without a text layer goes through OCR and counts as two pages; the API reports this as usage.ocr. The result counts too: every 2,000 characters of the JSON answer is a page and you never get fewer pages than that (usage.result_pages), so a wide item schema costs what you asked to extract.