The dedicated POST /v1/decisions endpoint turns text and images into typed answers using gpt-6-luna, the only supported model in this public beta. The available types are predicate, choice and score, covering condition checks, classification and scoring without framing every result as a conversational response.
OpenAI claims these uses run ten times faster than through the Responses API. That is a vendor statement about the specified workloads, not a measured speedup for every API application. The beta’s constrained output types and dedicated endpoint are the concrete change.
The October 6 changelog establishes availability, distinct from a possible September DevDay preannouncement. General availability is still expected later in the cited documentation. Belgium and EEA conditions are unspecified; access depends on the plan, rollout and administrator settings.
