One request to OpenAI's Decisions API can now cover 400 questions, following the service's move to general availability on Oct. 22 and letting developers turn text, images and structured data into defined results.
The new ceiling is twice the 200 questions allowed in beta. For each input, developers can ask several separate questions and specify whether the API should return a likelihood, a selection from preset options or a score.
The endpoint is aimed at workflows that need answers software can act on. A team can use these formatted results to sort its intake, with selected cases referred to staff for review.
GPT-6 Luna currently powers the endpoint. General availability brings native GPT-Live integration and prompt caching; inference can be assigned to a U.S. or EU region. Supported inputs include text, images and structured data.
Developers first received access through the public beta on Oct. 6. EconoTimes previously covered the Playground developers used to test the API.
OpenAI lists $0.10 for each million uncached input tokens and $0.01 for each million cached input tokens; output tokens have no fee. Earlier EconoTimes coverage noted that regional processing or longer context could increase input charges. The listed rates distinguish cached from uncached input, while output tokens carry no separate charge.
