OpenAI has launched its Decisions API in public beta, giving developers a new way to build applications that need fast, structured decisions instead of lengthy AI-generated responses.
The API was released in beta on October 6, 2026, following its earlier introduction in limited preview. OpenAI says the Decisions API can turn text and images into typed answers up to 10 times faster than using GPT-6 Luna through the Responses API.
The new service is designed for applications that need an AI system to make a specific choice quickly. Examples include selecting a model, routing a request to a tool, classifying content or determining the next action in an automated workflow.
The Decisions API is powered by GPT-6 Luna and uses the dedicated /v1/decisions endpoint. Developers can provide text or image-based input and ask the system to evaluate that information against predefined questions.
Rather than producing a traditional conversational answer, the API returns structured results that applications can use directly. This approach is intended to make AI-powered decision-making easier to integrate into high-volume software systems.
The API currently supports three main types of decisions: predicates, choices and scores. Predicates can estimate the probability that a statement is true, while choices allow an application to select from predefined options with confidence information.
Scores allow developers to evaluate an input against a numeric range or scoring system. These capabilities can be useful for applications that need consistent outputs rather than open-ended text.
Another notable feature is its ability to accept both text and images as input. OpenAI’s documentation states that the API supports user messages containing text and inline images, allowing developers to build decision workflows that work with visual information as well as written content.
The technology could have applications in areas such as automated classification, content moderation workflows, customer support routing, recommendation systems and software that needs to choose between different tools or actions.
For example, an application could use the API to determine whether a customer message relates to a particular issue, select an appropriate category or assign a numerical score based on predefined criteria.
The focus on structured outputs also means developers can design workflows where the AI’s response is immediately consumed by another part of an application.
OpenAI describes GPT-6 Luna as an efficient model for focused, high-volume tasks, making it a suitable foundation for a service designed around rapid decision-making.
The public beta means the Decisions API is now available for developers to test and integrate into their applications. As with other beta products, its capabilities and implementation details may continue to evolve as OpenAI gathers developer feedback.
The launch represents another step toward using AI not only to generate content but also to handle smaller, specialized decisions inside software systems.
For developers building fast AI-powered applications, the new API could provide a more direct way to add classification, selection and scoring capabilities without requiring a full conversational response for every request.
