OpenAI says its Decisions API is now available to all developers in public beta. In the company's launch posts, it says the API helps apps choose the right model, tool, or action in near real time.
OpenAI said the API is powered by GPT-6 Luna and accepts text and image inputs. In the same launch thread, the company said it supports three kinds of outputs: predicates, which estimate the probability that a statement is true; choices, which select from predefined options with confidence scores; and scores, which evaluate an input against a numeric range.
How OpenAI says developers are using it
OpenAI said developers have been using the API to route requests to the right model, tool, or agent; turn inputs into labels, rankings, and scores; analyze images and screenshots; flag risky tool calls; and categorize large datasets.
The company also posted a link inviting developers to try the Decisions API in public beta.
The main claim in the launch posts
OpenAI's launch copy says the Decisions API can make decisions up to 10x faster than GPT-6 Luna through the Responses API.
That speed comparison comes from OpenAI's own launch posts, and the supplied evidence does not independently verify how it performs across different tasks.
Some early social posts suggest interest, testing, and benchmarking around the API, but they do not independently confirm OpenAI's broader performance claims. Eric Rea of Podium said his company had alpha tested OpenAI's Decisions API to help agents choose the right workflow faster. Separately, a developer posting about HiringCafe said they had benchmarked the Decisions API against Jev for predicting how relevant a user query or resume is to a job description on a 1-10 scale, and claimed OpenAI was 2x more expensive and 5% to 10% worse on that specific task.
