Skip to content

Latest commit

 

History

History
61 lines (44 loc) · 1.71 KB

CALCULATOR.md

File metadata and controls

61 lines (44 loc) · 1.71 KB
Raw

Estimated Monthly Cost for Running the Accelerator

This page provides reference estimates for the cost of running the accelerator for one month.

1. Default GPT-RAG Zero Trust Deployment

To view the cost estimate for this scenario, visit the Azure Pricing Calculator. The link provides a breakdown of all solution components and their respective costs.

Usage Parameters Used for the Estimate:

Document Intelligence

  • Number of documents: 500
  • Pages per document: 20
  • Total pages: 10,000

CosmosDB

  • Data stored: 500 GB
  • Operations per second:
    • Creates: 20
    • Reads: 20
    • Updates: 20

Azure OpenAI

  • Number of user conversations per day: 100
  • Average interactions per conversation: 5
  • Total Azure GPT requests per month: 15,000

Token Calculation for Azure OpenAI

Prompt Tokens

  1. Triage: 450 tokens
  2. Answer: 221 tokens + Sources = 2,221 tokens
  3. Is Grounded: 99 tokens + Sources = 2,099 tokens

Total prompt tokens per 1,000 requests:

(2099 + 2221 + 450) * 15000 / 1000 = 71,550 tokens

Completion Tokens

  1. Is Grounded: 4 tokens
  2. Triage: 10 tokens
  3. Answer: 800 tokens

Total completion tokens per 1,000 requests:

(4 + 800 + 10) * 15000 / 1000 = 12,210 tokens

Note on Embeddings

For Ada embeddings, considering the volume of 500 documents used to generate embeddings, estimate an additional cost of approximately $100.

References

  1. CosmosDB Capacity Calculator
  2. Azure Pricing Calculator