Skip to content
TheChatArchiveThe Chat Archive
o4-mini-2025-04-16· OpenAIHugging Face LMArena Human Preference 140K

What is the cost to process 1M pages per day, with a lot of tables like in 80% of the…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Metallama-3.3-70b-instruct’s answer.
  1. User

    What is the cost to process 1M pages per day, with a lot of tables like in 80% of the pages, with AWS Textract?

  2. Assistant

    Here’s a back‐of‐the‐envelope cost estimate for processing 1 million pages/day with AWS Textract, assuming 80 % of your pages contain tables (and the other 20 % only need plain text extraction). We’ll use the asynchronous APIs, which are by far the most cost-effective at high volume.

    APIUse casePrice per 1 000 pagesPages/dayDaily cost
    AnalyzeDocument (tables & forms)80 % of pages (800 000)$15.00800 000$12 000.00
    DetectDocumentText (plain text OCR)20 % of pages (200 000)$1.00200 000$200.00
    TOTAL (per day)1 000 000$12 200

    If you simply ran AnalyzeDocument on all 1 million pages (even the ones without tables), it would cost
    1 000 × $15 = $15 000 per day.

    Monthly (30 days) at the hybrid rate:
    $12 200 × 30 ≃ $366 000

    Annual run-rate:
    $12 200 × 365 ≃ $4.45 million

    — plus whatever you pay for S3 storage, data-transfer, IAM roles, etc.

    Tips to reduce cost
    • Pre-filter pages with a lightweight model (e.g. Amazon Textract’s DetectDocumentText + custom logic) to only send truly “table-y” pages to AnalyzeDocument.
    • Consider Amazon A2I (Augmented AI) only for low-confidence extractions.
    • Batch smaller jobs in bulk (async) rather than synchronous calls.

    Feel free to let me know if you need a deeper breakdown or alternate scenarios (e.g. synchronous vs. asynchronous; volume discounts; or spot-price compute for pre-processing)!

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.