OpenAI's Decisions API is now in public beta, evaluating text, images, or both and returning typed answers about 10x faster than the Responses API. It provides probabilities, choices, or scores for tasks like content classification, request routing, and prioritization. The API is available in the Playground for experimentation before coding, with general availability expected in the coming weeks.
The Decisions API uses the POST /v1/decisions endpoint and currently supports only the gpt-6-luna model. A request includes the model, input, and questions, each with a type, instructions, and allowed choices or score levels. The response contains an answers array, and each question must have a unique name that the API echoes back.
Question types include predicate (checks a condition, returning a probability from 0 to 1), choice (selects one option from a fixed set), and score (rates input against ordered levels, returning a probability-weighted average of level indices). Use choice for unordered categories like departments, and score for ordered levels like severity.
For use cases requiring custom JSON schema generation or function calling, OpenAI recommends Structured Outputs with the Responses API instead. Images must be inline base64 data URLs; hosted URLs and file_id inputs are not supported. The API combines input_text and input_image parts in a user message for multimodal evaluation.
Source: Hacker News · Summarized by HeadlinesBriefing