IT Home News on September 30th: During today's 2026 Developer Day event, OpenAI announced the launch of Decisions API. This is a new API designed for real-time, low-latency classification and routing scenarios, capable of returning results in about 150 milliseconds, which is approximately 10 times faster than using Luna through conventional API.
In terms of positioning, Decisions API is designed for "quick, single-instance decision-making," focusing the "intelligence" of the model (the current small model Luna based on OpenAI) on a set of user-defined questions and a limited set of answers, returning structured decision results with extremely low latency.
Among the current models of this type, the main competing model comes from TypeSafe's Jev. However, Decisions API focuses on transforming the model's intelligence into rapid decision-making capabilities.
Developers provide context (text or images) and predefine "questions + optional answers." API returns the selected answer along with structured information such as confidence level, which is used for subsequent business logic.
In terms of typical use cases, Decisions API is more suitable for the following scenarios:
- Customer Service/Ticket System: Automatically determines the type of issue, priority, and the team to be assigned based on the user's description.
- Content review and policy selection: Quickly choose from several predefined policies (for example: allow / rate limit / manual review).
- Agent "Decision Nodes" in Layout: In multi-step workflows, use Decisions API to determine which tool/sub-agent to call next, rather than letting large models act freely.
- Real-time business rule assistance: On top of the existing rule engine, Decisions API is used to handle ambiguous situations, providing suggested decisions within a limited range of options, followed by the final ruling made by the rule engine.
