Build more responsive apps with the Decisions API

Built on GPT-6 Luna, the endpoint handles bounded questions up to 10x faster than the Responses API, charging $0.10 per million input tokens with no output fees.
Key points
- Supports three structured output types: probability predicates, predefined choices, and rubric scores.
- Accepts text, images, and native JSON inputs, handling up to 400 questions per request.
- Cached input costs $0.01 per million tokens, with inference residency options in the US and EU.
- Integrates natively with GPT-Live to trigger actions and animations during real-time voice interactions.
Why it matters
Developers building AI agents and classification pipelines can execute high-frequency routing and categorization decisions with significantly lower latency and cost.
Sources
- Build more responsive apps with the Decisions API (OpenAI, Oct 22, 2026, 7:47 PM)