Azure Openai gpt5.6

Question

Does Azure OpenAI GPT‑5.6 support cache read/write and tiered pricing? If possible, could you provide a prompt that can generate a cURL command to create a cache write? What is the price for cached_tokens?

Answer 1

Hello @51155945

Thank you for your question.

At this time, I could not find any official Microsoft documentation confirming whether GPT-5.6 supports prompt caching (cache read/write), cached token pricing, or tiered pricing models. The availability of these features and their pricing are determined by the model and deployment configuration, and are documented only when officially released.

For the most accurate and up-to-date information, please refer to the Azure OpenAI pricing and model documentation, and verify whether GPT-5.6 specifically lists support for prompt caching and cached token billing.

If your goal is to create a cached request, the API request format generally depends on the prompt caching feature being supported by the deployed model. Please share the region and deployment type you are using (Azure OpenAI or Azure AI Foundry), and we can help review the supported capabilities further.

Thank you.

Answer 2

Hi! A useful way to separate this is that prompt caching is service-managed; you do not normally send a separate cache-write request. Send the same stable prompt prefix in later calls and check the usage details for cached tokens. Whether GPT-5.6, a particular deployment type, and cached-token pricing are available depends on the Azure region/API version and can change during rollout, so the model catalog and current Azure OpenAI pricing page are the source of truth for that deployment. I would first run two identical curl requests against the exact Azure deployment and compare the second response's cached-token usage. If it stays at zero, confirm the model/version and deployment tier support caching before changing the application.