Understand
Introducing the Decisions API for Fast Model Responses
Overview
OpenAI introduces the Decisions API, powered by GPT-6 Luna, designed to allow models to respond in a fraction of a second by focusing on a specific set of choices for text or image inputs. The API enables fast application actions, such as routing support tickets, controlling car navigation from video frames, driving character expressions in voice conversations, and guiding robotics.
Key points
- The Decisions API runs on GPT-6 Luna and responds in a fraction of a second (under 100 milliseconds on the server side).
- Users provide text or image inputs, define a question, and specify possible answers to obtain fast classifications and actions.
- The API maintains capabilities like image understanding, broad language support, and safety protections while operating nearly 10 times faster than reasoning models.
- Demonstrations include automatically classifying support tickets or sales leads, steering a car in a driving game, controlling animated character facial expressions, and guiding a programmable robot named Lavender.
- It pairs effectively with voice modalities like GPT-Live-1 and can be integrated into interactive hardware and robotics experiences.