OpenAI’s Decisions API leaves beta

What changed

Agents make many tiny calls, like where to route a request or whether to retry, and the waiting adds up. The Decisions API is OpenAI’s dedicated endpoint for exactly those calls.

How it works

You define a question and its allowed answers, then give it context as text, images or native JSON. It returns one of three answer types:

• Predicate: the probability, from 0 to 1, that a condition is true, such as whether a product has visible damage.
• Choice: one of your predefined options, such as continue, retry or ask for clarification, with probabilities and confidence.
• Score: a rating against ordered levels in a rubric, such as how severe the damage is.

You can ask up to 400 questions per request, twice the earlier limit.

How teams use it

In OpenAI’s announcement, Datadog uses it to score the risk of agent-generated commands before running them, Box classifies business documents and routes them to the right next step, and Sierra uses it to choose the next action during parts of live customer conversations.

New since the beta

A native GPT‑Live integration for voice apps, prompt caching for reused context, and inference residency in the US and EU. Cached input costs $0.01 per million tokens, and Zero Data Retention is available to eligible customers.

Try it

OpenAI suggests starting with one decision your app makes often: define the question and possible answers, then test it on real examples in the Decisions Playground or with the developer guide.