Azure Document Intelligence Pre-built Models โ All 15 Models from Receipts to Contracts
prebuilt-read, prebuilt-receipt, prebuilt-contract, prebuilt-invoice โ change model name, get structured data
General OCR vs Azure Specialized Models
General OCR: image โ one text blob. Parse it yourself.
Azure specialized: image โ structured JSON with named fields. No parsing needed.
All Pre-built Models
General: prebuilt-read (OCR), prebuilt-layout (structure)
Business: invoice, receipt, contract, creditCard, bankStatement, check, payStub, marriageCertificate, mortgageClosingDisclosure
ID: idDocument (passport/license), healthInsuranceCard.us
Tax (US): tax.us.w2, tax.us.1098, tax.us.1099
Code โ Just Change Model Name
# Receipt
result = client.begin_analyze_document("prebuilt-receipt", body=f).result()
# โ {"MerchantName": "Starbucks", "Total": 5.50}
# Contract
result = client.begin_analyze_document("prebuilt-contract", body=f).result()
# โ {"Parties": [...], "Jurisdictions": [...]}
# Passport
result = client.begin_analyze_document("prebuilt-idDocument", body=f).result()
# โ {"FirstName": "John", "DateOfBirth": "1990-01-01"}
Same code structure. Different model name โ different structured output.
Pricing
prebuilt-read: ~$1.50/1K pages. Specialized models: ~$10/1K pages. The structuring costs extra.
Custom Models
No matching pre-built? Train your own in Azure Portal with 5+ labeled samples. No code needed.
Key Concepts
prebuilt-read โ general OCR ($1.50/1K). Text + position coordinates
prebuilt-receipt โ receipt specialized ($10/1K). Auto-extract store, date, amount
prebuilt-contract โ contract specialized ($10/1K). Parties, jurisdiction, clauses
prebuilt-idDocument โ passport/license ($10/1K). Name, DOB, document number
Custom model โ label 5+ samples in Azure Portal to train your own