OpenAI has opened a Playground for developers to test the Decisions API, which is in public beta and returns answers to defined questions about text and images before teams add it to applications.
Developers can frame a query in three ways. Depending on the task, developers can ask the API to estimate whether a condition holds, choose among supplied options or rate an input on a defined scale. A single request can include multiple independent questions about the same input. Those formats let developers choose whether a task needs a probability, a selection or a rating.
The API is designed for work such as classifying content, routing requests and prioritizing tasks. OpenAI’s Playground lets developers try different questions against text and image inputs before connecting the API to an application. Developers define the condition, options or scoring levels, then check whether the returned answer fits the decision their workflow requires.
At the base rate, developers pay $0.10 per million input tokens. Regional processing or a longer context can raise that charge; Decisions outputs carry no token fee. API customers pay the input charges and any applicable premiums, so costs vary with usage and processing choices.
OpenAI released the API in public beta on Oct. 6. It currently supports only gpt-6-luna. OpenAI says the service can return answers up to 10 times faster than GPT-6 Luna through the Responses API.
The company expects general availability in the coming weeks but has not announced a date.
