Azure AI Vision vs Document Intelligence Pricing for Invoice OCR

Short answer: for plain OCR, Azure AI Vision Read and Azure Document Intelligence Read cost the same, $1.50 per 1,000 pages, dropping to $0.60 after 1 million pages a month. Neither returns invoice fields. To get vendor, invoice number, totals and line items on Azure you pay for the Document Intelligence prebuilt invoice model at $10 per 1,000 pages, which only exists on the Document Intelligence side. For invoice work, Document Intelligence is the service to price. Vision is only the cheaper buy if you commit to 500,000 pages a month or more of text-only OCR.

Azure shops usually hit this question in the first week of an accounts payable project. Someone finds the Vision OCR quickstart, someone else finds the Document Intelligence Studio, and both look like they read invoices. They share the same OCR engine and the same list price, so the choice looks arbitrary. It isn't. This article prices both services for a US AP team, using the rates in the Azure Retail Prices API for East US. You can drop one of your own invoices into the tool at the top of this page to see what a finished extraction looks like while you read.

What each Azure service actually sells

Azure AI Vision (formerly Computer Vision) is an image service. Its Read feature pulls printed and handwritten text from photos and PDFs, and the same resource also does tagging, captions, object detection and smart cropping. Read sits in the Image Analysis Group 2 meter.

Azure Document Intelligence (formerly Form Recognizer) is a document service. It runs a document-tuned version of Read, and adds Layout (tables, sections, reading order), prebuilt models for invoices, receipts and ID documents, and custom extraction models you train on your own forms.

Microsoft's own OCR documentation draws the line clearly. It recommends Vision Read (version 4.0) for general, in-the-wild images such as labels, street signs and posters, and Document Intelligence Read for scanned and digital documents. Its training material adds that the Document Intelligence Read model runs at a higher resolution than the Vision version. The older Vision v3.2 Read API gets no further updates.

Azure AI Vision vs Document Intelligence pricing, side by side

The Read rows are identical, which is why so many teams assume the two services are interchangeable. Everything below the Read rows is where they split: Vision stops at text, and Document Intelligence keeps going into structure and fields. The full Vision rate card, including its cheaper Group 1 meters, is on our Azure OCR pricing page.

Is Azure AI Vision cheaper than Document Intelligence?

Only for text, and only at volume. On pay as you go the two cost exactly the same for Read. The one place Vision wins is its prepaid Read blocks: 500,000 pages for $375 a month, which is $0.75 per 1,000, half the pay-as-you-go rate. At 2 million pages the block costs $1,200 against $2,100 on pay as you go.

For an AP department those savings are real but small in dollars. A team processing 3,000 invoices a month at 1.4 pages each sends about 4,200 pages, and Read costs $6.30 on either service. Nobody picks an OCR vendor over six dollars. The decision turns on what you do after the text comes back.

What invoices cost on each service

Here is what 5,000 single-page vendor invoices a month cost, with the work that remains after the API call.

The Vision route looks cheapest until you count the parser. Invoices put the total in a different place on every vendor's layout, and line-item tables break across pages, wrap descriptions onto two lines and mix quantities with unit prices. Writing and maintaining that logic takes far more than the $42.50 a month you save over the prebuilt model. The Content Understanding row is there because Microsoft now presents it as the successor for document extraction, but its field meter costs more than double the prebuilt invoice model and is only sold in West US among US regions. We cover that in detail on the Azure Content Understanding pricing page.

Which Azure service should I use for invoice OCR?

Use Document Intelligence. It is the only Azure service with an invoice model, its Read engine is tuned for documents, and the Read rate is the same as Vision's. The prebuilt invoice model at $10 per 1,000 pages returns typed header fields and line items, and commitment tiers bring that down to $9.50 at 20,000 pages a month and $7.50 at 1 million. Our Azure Document Intelligence pricing page lays out every tier, including the container options.

Use Vision Read only when invoices are not really the job: indexing scanned archives for search, reading text from product or shelf photos, or feeding page text into a model you already run. If you are weighing the whole OCR market rather than just Azure, our comparison of Mistral OCR vs Azure Document Intelligence shows where a newer OCR model undercuts Microsoft on price.

When neither Azure service is the right buy

Both Azure services are developer building blocks. Even the prebuilt invoice model hands back JSON with confidence scores, not rows in your accounts payable ledger. Someone still has to provision the resource, handle the asynchronous polling, map fields to your chart of accounts, build a review step for the low-confidence values and write the export. After that, the extracted rows usually have to move into an ERP or a data warehouse, and many teams end up adding a data integration platform just to keep those feeds running.

For a finance team without spare engineers, that build is the real cost of the Azure route, and it usually outweighs years of API charges. A finished extractor reads the header and each line, lets a person confirm the result on screen, and exports Excel, CSV or JSON at a flat monthly price, with no Azure subscription, keys or polling loop. Upload one of your own invoices in the tool at the top of this page and compare the output with a Document Intelligence JSON response before you decide which one your team should maintain.

Frequently asked questions

Is Azure AI Vision the same as Document Intelligence?

No. Both use Microsoft's Read OCR engine at the same $1.50 per 1,000 pages, but Vision is an image service that stops at text, while Document Intelligence adds layout, prebuilt invoice and receipt models, and custom extraction. Microsoft recommends Document Intelligence Read for scanned and digital documents.

Can Azure AI Vision extract invoice data?

It extracts the text on an invoice, not the invoice fields. Vision Read returns lines and words with coordinates and does not label the vendor, invoice number, due date, total or line items. For fields on Azure you need the Document Intelligence prebuilt invoice model, priced at $10 per 1,000 pages.

How much does Azure Document Intelligence cost for invoices?

The prebuilt invoice model costs $10 per 1,000 pages on pay as you go, so 5,000 single-page invoices cost $50 a month. Commitment tiers start at $190 a month for 20,000 pages, an effective $9.50 per 1,000, and fall to $7.50 per 1,000 at the 1 million page tier.

What happened to Azure Computer Vision and Form Recognizer?

They were renamed. Computer Vision became Azure AI Vision and Form Recognizer became Azure AI Document Intelligence, and both now sit under Microsoft's Foundry Tools branding. The APIs, meters and prices carried over, so older tutorials that use the previous names still describe the same services.