Summary
OpenAI introduces the Decisions API, a new developer endpoint built for situations where an app needs to make a smart choice almost instantly. Running on GPT-6 Luna, it takes a text or image input along with a question and a fixed set of possible answers, then returns a decision in a fraction of a second. By narrowing the model to a handful of choices, OpenAI says it is nearly 10 times faster while retaining image understanding, broad language support, and safety protections.
The launch video walks through several demos. On the text side, the API classifies loose user input, extracts form fields, and routes support tickets and inbound sales requests, with server-side responses under 100 milliseconds. On the vision side, it steers a car in a simple game by choosing lanes from video frames, far faster than a reasoning model could.
The presenters also pair the API with GPT-Live-1 for voice, where it selects an animated character’s expressions during conversation. In a robotics demo, a preview unit of Hugging Face’s programmable Microduck robot uses it to decide where to look, following an apple, a piece of fruit, or whichever object is more fun to play with. OpenAI positions it as a complement to more capable models for sophisticated computer-use scenarios.
📺 Source: OpenAI · Published October 06, 2026
🏷️ Format: Keynote Launch
