Product Updates

OpenAI Decisions API: Public Beta, Endpoint & Use Cases

OpenAI’s Decisions API is in public beta with GPT-6 Luna. See the /v1/decisions endpoint, text and image inputs, typed answers, and use cases.

By Authority AI Tools Editorial Team•2026-10-09•10 min read

Last reviewed: 2026-10-09

AATET

Authority AI Tools Editorial Team

Editorial Team

The Authority AI Tools editorial team maintains this directory using vendor documentation, dated source checks, product changelogs, and clearly identified hands-on observations where available.

OpenAI announced the Decisions API at DevDay on September 29, describing it as a way to make bounded decisions from user-defined questions and a finite set of answers. Developers can provide context as text or images and use the result to classify content, route a request, or choose an agent’s next action.

Public beta: endpoint, model, and answer types

OpenAI’s API documentation now lists Decisions in public beta, following the limited preview announced at DevDay. The guide names GPT-6 Luna as the only currently available model and documents the dedicated POST /v1/decisions endpoint. Requests can include text, images, or both as context for one or more named questions.

The API returns typed answers in three forms:

Predicate: estimate the probability that a stated condition is true.
Choice: select one of the developer-supplied options, with probability and confidence information.
Score: evaluate an input against an ordered rubric and return a score with per-level probabilities.

OpenAI says general availability is expected in the coming weeks. Beta access, the request schema, model support, and pricing can change; check the live API guide before implementing a production integration.

Keep the question and answer type explicit

The API is designed for bounded decisions such as classifying content, routing a request, or prioritizing work. Use a choice question for unordered categories, a predicate for a condition, and a score when the options have a meaningful order. OpenAI recommends making choices distinct and defining clear criteria for score levels.