OpenAI has announced a Decisions API, a service that lets an app pass a choice to a model instead of coding the rules itself. The company’s developer account on X says it runs on GPT-6 Luna.
Developers define the questions and the possible answers, and the model picks one. OpenAI’s post names three uses: classifying content, routing requests, and choosing an agent’s next action. In plain terms, this is sorting work that many products now handle with hand-built rules, or with a general chatbot prompted to reply with a single label.
The captured post is short. It stops mid-sentence at the point where it would describe availability, so it confirms no price, access terms, launch date or customer. Nothing in it offers independent test results either, and any claim about speed or accuracy would be OpenAI’s own.
The market angle is straightforward. A dedicated endpoint for structured choices puts OpenAI in competition with the narrow classification models that smaller vendors and open-source projects already sell for the same jobs.
Teams running routing or moderation logic should hold off on any migration plans until OpenAI publishes pricing and access terms.
Announced by OpenAI’s developer account on X, as republished in a Thread Reader thread; the captured post carries no date.
