Exact billable input token counting before Azure OpenAI Responses API request

Luey 0 Reputation points
2026-10-11T12:27:37.97+00:00

We use Azure OpenAI gpt-5-mini, version 2025-08-07, SKU DataZoneStandard, area EUR, Sweden Central, via the Responses API with Structured Outputs.

We need an authoritative answer for strict production cost control.

Is there a Microsoft supported method or API to determine the complete billable input token count before sending the request?

Does the billable input_tokens count include the JSON schema, provider framing, hidden system framing, or any other tokens not directly present in the serialized request body?

If exact precounting is not available, does Microsoft provide a deterministic upper bound or formula for the maximum billable input tokens for a known request including Structured Outputs?

Is usage.input_tokens returned by the Responses API the authoritative billable input token count for this Azure deployment?

Does any supported token counting endpoint itself incur Azure OpenAI model charges?

We need the answer specifically for gpt-5-mini, version 2025-08-07, DataZoneStandard, Sweden Central, Responses API.

Cost Management
Cost Management

A Microsoft offering that enables tracking of cloud usage and expenditures for Azure and other cloud providers.

0 comments

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.