About the news
The API is built for the many small decisions that applications and agents make during a workflow — classifying an input, routing a request, or choosing the next action. OpenAI says these decisions are simple on their own but add up in latency and cost at scale.
The Decisions API runs on GPT-6 Luna and focuses the model on questions with bounded answers. OpenAI reports it can return decisions up to 10 times faster than using the same model through the Responses API. Developers define a question and the allowed answers, then pass text, images, or structured data. The API returns one of three types of answer: a probability that a statement is true, a choice from a fixed set of options, or a score against a rubric.
What changed since the public beta:
- Native GPT-Live integration so voice apps can react while a person is still speaking
- Prompt caching for reused context
- Native JSON inputs so structured app data can be passed directly
- Images can be included in individual question instructions
- Up to 400 questions per request (double the previous limit)
- Inference residency in the US and EU
Pricing
Pricing is $0.10 per million input tokens, or $0.01 per million for cached input. There are no charges for output tokens. Developers pay only for the input.
