OpenAI's Decisions API is now open to all developers, charging only for input at $0.10 per million tokens.

On Thursday, Oct 22, OpenAI made its Decisions API generally available, after a public beta that started earlier this month (OpenAI). It's built for the many small yes-or-no and pick-one questions apps and agents face all day, like how to label a message, where to route a request or what to do next. You write a question and its allowed answers, pass in text, images or JSON data, and get back one of three answer types: a probability that something is true, a choice from your list, or a score on a scale you define. OpenAI says it runs on GPT-6 Luna and answers up to 10x faster than asking GPT-6 Luna the same thing through its regular Responses API.

Since the beta, OpenAI has added prompt caching, a hookup to its GPT-Live voice model, data processing inside the US or EU, images in the question instructions, and up to 400 questions per request, double the old limit. It costs $0.10 per million input tokens and $0.01 per million cached input tokens, with no charge for output.

Companies quoted in the post report gains from their own tests: Sierra says it picks the next step in live customer conversations in under 280 milliseconds at 90% lower cost per call, and Speak says reaction time in its language lessons fell from 1.4 seconds to 0.13 seconds (OpenAI Developers on X).

What to weigh: if your app or agent makes lots of simple calls, like sorting messages or choosing the next step, this can be much cheaper and faster than a full chat model. It only answers bounded questions, so anything that needs written output still needs a regular model.