Global Tech News For people figuring out what to build with AI.
AI

OpenAI adds an endpoint for structured text and image decisions

A large call centre in Lakeland, Florida, photographed in 2006; illustrative context, not an OpenAI interface or deployment.

OpenAI opened its Decisions API in public beta on Oct. 6, giving developers a dedicated endpoint that turns text or images into structured answers such as a probability, a selection from supplied options or a score against a rubric. It uses GPT-6 Luna; this is a new API route, not a new model.

An application sends evidence and one or more independent questions. For a support message, for example, it could ask whether the customer reports a damaged item, choose a team from a fixed list, and rate urgency against ordered levels. The API returns answers in structured fields. The application still decides whether a result should trigger automatic routing, human review or another action.

The endpoint offers three answer types. A predicate estimates the probability that a condition is true. A choice selects one of the options supplied by the developer. A score evaluates the input against ordered levels and returns an average that gives more weight to levels the model considers more likely. OpenAI says developers should use the Responses API instead when they need a custom JSON schema, a written explanation or function calls.

OpenAI says the new route is about 10 times faster than the Responses API, but the documentation does not specify test inputs or other conditions for that comparison. The listed price is $0.10 per million input tokens, with no charge for output tokens or cache reads and writes; regional processing premiums and long-context multipliers still apply.

The API remains in public beta, with GPT-6 Luna the only supported model in the documentation. Its probability and confidence fields are model outputs, not a ready-made business policy: OpenAI advises application teams to set decision thresholds using labeled examples and the relative cost of false positives and false negatives.

Microsoft announced its separate Microsoft-Decision-1 model three days later for structured routing and scoring. Microsoft describes a standalone model, while OpenAI has packaged similar tasks as an endpoint on GPT-6 Luna. GTN previously reported on Microsoft's release.

Sources