Introducing the Decisions API
Summary
OpenAI's Decisions API delivers AI responses in under 100ms by constraining models to binary/multiple-choice outputs, enabling 10x faster decision-making for real-time applications like routing, classification, and robotics without sacrificing vision or language capabilities.
Key Takeaways
- Decisions API responds in <100ms by restricting model output to predefined choices, making it ~10x faster than standard models while maintaining image understanding and safety protections.
- Concrete use cases include text classification (routing support tickets/sales leads), real-time game steering, dynamic character expressions paired with voice, and robotics object tracking—showing 4 distinct application patterns.
- Pairs effectively with GPT-4o for voice and GPT-Live-1 for live interactions, enabling hybrid architectures where reasoning models handle complex tasks while Decisions API handles real-time micro-decisions.
- Vision-based frame analysis enables subsecond decisions on visual input—demonstrated by steering game car and robot head tracking—creating opportunities for computer vision applications requiring instant feedback loops.
- Ambiguous queries work effectively (e.g., 'follow the fruit' across multiple objects), suggesting the API maintains semantic understanding despite speed optimization, reducing need for rigid input formatting.
Related topics
Transcript Excerpt
Sometimes your app needs to make a smart decision really, really fast. Routing a request, classifying an image, or deciding what an agent should do next. Today, we're introducing the Decisions API. It lets the model respond in a fraction of a second, so interacting with it feels pretty much instant. The Decisions API runs on GPT-6 Luna. It's very simple. You give it a text or an image, you define the question and the possible answers, and you use the result in your app. By focusing the model on just a few choices, we can make it nearly 10 times faster while keeping capabilities like image understanding, broad language support, and safety protection. But let's see that in action with Charlie, starting with text input. Thanks Romain. We're already finding a wide variety of applications that …