OpenAI has released the Decisions API in public beta, with general availability expected in the coming weeks. According to OpenAI’s documentation, the API evaluates text, images, or both and returns typed answers approximately 10 times faster than the Responses API.
The Decisions API supports three types of queries: predicate questions that check conditions and return a probability between 0 and 1; choice questions that select one option from a fixed set; and score questions that rate inputs against ordered levels and return a probability-weighted average.
Use cases include classifying content, routing requests, and prioritizing work within applications. For example, a predicate question could inspect a product photo for visible damage such as cracks, tears, or dents. A choice question might assign customer support messages to departments, while a score question could evaluate issue severity across ordered levels.
Currently, gpt-6-luna is the only available model. Developers access the API through a dedicated POST /v1/decisions endpoint. OpenAI has provided code examples in Python, JavaScript, Go, Ruby, and Java. To use the SDK examples, developers need Python 3.26.0 or later, JavaScript 7.30.0 or later, Go 3.73.0 or later, Ruby 0.101.0 or later, or Java 4.78.0 or later.
Requests to the Decisions API contain three parts: a model specification, input evidence as text or messages containing text and images, and questions defining what to evaluate. The response includes an answers array, with each question identified by a unique name that appears in the response.
OpenAI distinguishes the Decisions API from other features: use Structured Outputs with the Responses API when generating objects that follow a custom JSON schema, and use function calling when the model needs to request a tool call with arguments. The Decisions API is specifically designed for applications that need one of the three answer types it provides.
Key facts
- The Decisions API evaluates content 10x faster than the Responses API
- Three question types are supported: predicate (probability), choice (selection), and score (weighted average rating)
- Currently only gpt-6-luna model is available
- General availability is expected in the coming weeks
- API is accessed via a dedicated POST /v1/decisions endpoint