AI news · October 8, 2026
OpenAI adds a Decisions API for typed choices and scores
The Decisions API is aimed at small, repeated judgments rather than open-ended replies. OpenAI says it is 10x faster than the Responses API and charges $0.10 per million input tokens.
- claimed speed over Responses API
- 10x
- input price per million tokens
- $0.10
- model in public beta
- 1
OpenAI opened the Decisions API in public beta on October 6. The dedicated endpoint accepts text or images and returns a typed predicate, choice, or score, which gives an application a direct answer for a narrow decision instead of a long generated response. The beta currently supports GPT-6 Luna. OpenAI says the endpoint is 10x faster than the Responses API and costs $0.10 per million input tokens, with no output-token charge.
This is a useful API shape for routing, filtering, form checks, and simple eligibility decisions. It also creates a clean place to test the model against labeled examples before adding it to a larger agent. The stated speed and price are vendor claims, so a real test must include retries, image inputs, false positives, and the cost of acting on a wrong decision. A typed answer is easier to log, but it is not automatically a safe answer.
What you can do with it
Try the Decisions API guide with a small labeled set of real inputs. Define the allowed outputs before calling the model, then compare accuracy, latency, and retry rate with your current Responses API flow. Keep a human review path for decisions that can deny access, spend money, or change a customer record.
Our take
A narrow decision endpoint is more useful than another general chat surface for teams that already know the output shape. The value will show up in simpler application code and lower latency. OpenAI still needs to prove the claimed speed on production inputs, especially images and long context.
Links Try today API surface Announcement
Source: OpenAI ↗ — Made With Models writes the brief; the reporting is theirs.