OpenAI announced that the Decisions API is now generally available, using the GPT‑6 Luna model to provide focused decision outputs for questions with bounded answers. The company says the Decisions API can make near real-time decisions from text, images and structured inputs, and in some cases runs up to 10× faster than using GPT‑6 Luna through the Responses API.
Purpose and capabilities
The Decisions API is designed to handle many small, repeated decisions across applications and agents—such as classifying inputs, routing requests, or choosing the next action in a workflow—quickly and at low cost. Developers define a question and allowed answers, then supply context as text, images, or native JSON. The API returns one of three supported answer types that applications can use to determine subsequent behavior:
- probability: an estimate between 0 and 1 that the condition is true,
- choice: one of the developer-supplied values (the response also includes probabilities and confidence),
- score: the probability-weighted average of indexed rubric levels (useful for severity or graded scales).
Features added since public beta
Since the public beta, OpenAI has added:
- native GPT‑Live integration for making decisions during live conversations,
- prompt caching for reused contexts,
- inference residency (regional processing) in the United States and the European Union,
- support for native JSON inputs and including images in question instructions,
- an increased per-request question limit to 400 (double the previous limit),
- continued latency improvements and additional capacity in several regions in Asia.
Early internal and customer use
OpenAI teams are already applying Decisions across product and business workflows: the Dots team uses it to reduce response times and improve interactivity; the Personal Finance team uses it to categorize transaction descriptions; and the Communications team uses it to classify coverage by themes and to score whether content reinforces or challenges key narratives, helping to flag items that need closer review.
Customers have used the API for a variety of tasks, from accurately attributing AI costs to making language-learning experiences more interactive.
Live voice and avatar scenarios
With GPT‑Live integration, Decisions can power real-time choices in voice applications. Examples include a support agent deciding to ask a clarifying question, route or escalate a request, or an avatar adjusting facial expressions based on a user’s words to create a more interactive experience. Live sessions emit decision events that applications can read to animate reactions.
How to start and pricing
Developers begin by selecting a recurring decision their application makes often, defining the question, possible answers, and necessary context, and testing on representative examples using the Decisions Playground or the developer guide. The Decisions API is available today via a dedicated Decisions endpoint using GPT‑6 Luna.
Pricing details provided in the announcement:
- Decisions API input tokens: $0.10 per million tokens,
- Cached input tokens: $0.01 per million tokens (pricing varies for 30-minute and 12-hour cached writes),
- Only input tokens are charged—there are no output-token fees.
Regional processing is available in the US and EU, and Zero Data Retention is offered to eligible customers. OpenAI also directs developers to documentation for current rates, caching details and data controls.
The Decisions API targets fast, structured decision-making for a wide range of use cases where many small, high-frequency decisions drive user experience or operational workflows.
