Do you know the price of Claude Opus 4-5

Pranav Gupta 1 Reputation point
2025-12-01T23:21:38.0666667+00:00

Do you know the price of Claude Opus 4-5, I can see this and according to this, it's $5 for 1M input and $25 for output that seems to be cheap for opus. Also If i deploy it on ai foundry this will be deployed as base model and we won't we charged for anything else other then the API being called.
User's image

Foundry Tools
Foundry Tools

Formerly known as Azure AI Services or Azure Cognitive Services is a unified collection of prebuilt AI capabilities within the Microsoft Foundry platform


1 answer

Sort by: Most helpful
  1. Anonymous
    2025-12-02T01:36:04.2833333+00:00

    Hi Pranav Gupta

    Welcome to Microsoft Q&A and Thanks for sharing the details.

    You’re reading the pricing correctly the image confirms the pay‑as‑you‑go rates for Claude Opus 4‑5 on Azure AI Foundry:

    • Model input: $0.005 per 1,000 tokens
    • Model output: $0.025 per 1,000 tokens

    Cache-related costs (optional optimizations):

    Write tokens: $0.00625 per 1,000

    • Hit tokens: $0.0005 per 1,000
    • 1‑hour cache write: $0.01 per 1,000

    That translates roughly to $5 per million input tokens and $25 per million output tokens, which matches what you mentioned.

    Deployment on Azure AI Foundry

    • When you deploy Claude Opus 4‑5 as a base model, there’s no extra hosting fee—you only pay for API inference calls (token usage and any optional caching).
    • No separate Anthropic contract; billing flows through your Azure subscription.
    • Startup credits don't apply. It is only available for enterprise and MCA-E subscriptions.

    Deploy and use Claude models in Microsoft Foundry - Microsoft Foundry | Microsoft Learn

    I Hope this helps. Do let me know if you have any further queries.

    Thank you!
    UPDATE: Updated Azure Startup Billing and Usage for Claude Model.

    Was this answer helpful?


Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.