OpenAI makes Decisions API generally available
OpenAI has moved its Decisions API, built on GPT-6 Luna, to general availability, offering developers bounded-answer classification and routing up to 10x faster than Luna via the Responses API, priced on input tokens only.
OpenAI announced that its Decisions API is out of public beta and generally available. The API narrows GPT-6 Luna to questions that have bounded answers, the kind of small call apps and agents make constantly, such as classifying an input, routing a request or choosing a next action. OpenAI says it returns decisions from text, images and structured inputs up to 10x faster than calling GPT-6 Luna through the Responses API.
Developers write a question, list the allowed answers and pass in context. The API supports three answer types:
- predicate: the probability, from 0 to 1, that a condition is true
- choice: one of the developer's predefined options, plus probabilities and confidence
- score: a probability-weighted rating against ordered levels in a rubric
Changes since the beta include a native integration with GPT-Live, OpenAI's voice product, which sends decision events during live conversations. OpenAI's examples are support agents deciding when to escalate and avatars changing their facial expressions. Also new are prompt caching, inference residency in the US and EU, native JSON inputs, images in question instructions and up to 400 questions per request, double the old limit. OpenAI has also added capacity in several regions in Asia.
The API costs $0.10 per million input tokens and $0.01 per million cached input tokens. There is no charge for output tokens. Zero Data Retention is available to eligible customers.
Internally, OpenAI teams use the API to categorize personal-finance transactions and to classify press coverage by theme. OpenAI did not name any external customers or publish benchmark details behind the latency claim.