Hi Chris,
Azure Document Intelligence uses a pay-as-you-go model based on the number of pages you process. The pricing tiers differ based on the features you need, which is why understanding the modules is key.
The service is structured around two main types of models:
Prebuilt Models: These are ready-to-use for common documents like invoices, receipts, ID cards, and general layouts. You are charged per page.
Custom Models: You can train these on your own document types for higher accuracy. There's a cost for training the model and a separate, usually lower, cost for analyzing pages with it.
For your specific case of 50,000 pages per year, here's my recommendation:
I suggest using the General Document prebuilt model (part of the "Read" and "Layout" functionality). It's a great, cost-effective starting point for most text and structure extraction needs.
At the S0 pricing tier, this model costs $1.25 per 100 pages.
For 50,000 pages, your estimated annual cost would be: (50,000 / 100) * $1.25 = $625.
There is no separate "subscription" fee; you would just pay for what you use, and this usage would be billed monthly. If your documents are more complex (like invoices where you need to extract specific fields like InvoiceTotal), you would use the specialized prebuilt Invoice model, which is priced higher.
If you have a more consistent, predictable monthly usage, you could consider an Azure savings plan for compute to reduce costs, but for your volume, pay-as-you-go is likely the most straightforward.
Please click Accept Answer if this was helpful.
rgds,
Alex