OpenAI launches Decisions API with up to 400 questions per request
3-line summary
- OpenAI made its Decisions API generally available on October 22.
- A request can now contain up to 400 questions, twice the beta limit.
- GPT-6 Luna evaluates text, images and JSON, returning probabilities, choices or scores.
OpenAI made the Decisions API generally available on October 22. It targets bounded decisions, such as routing a customer request to a department.
Since beta, the question limit has increased from 200 to 400 per request. OpenAI also added prompt caching and a native GPT-Live integration.
The developer guide currently lists GPT-6 Luna as the supported model. It offers predicate probabilities, choices from supplied options, and rubric-based scores, using text, images or JSON as evidence.
Developers still need to check decisions against real cases. The guide recommends labeled application examples to set thresholds, weighing the costs of false positives and false negatives.
Why it matters
Teams gain another option for routing and classification. Adoption should depend on real workflow errors as well as response time.
Who it's for
- Developers
- Compare one classification task with your existing approach using labeled cases.
- Office workers
- Decide which requests should reach a human reviewer.
Sources
- Build more responsive apps with the Decisions API — OpenAI
- Decisions — developer guide — OpenAI API documentation
