OpenAI opens its Decisions API to all developers

OpenAI made its Decisions API generally available on Thursday, a service that answers bounded questions (a yes/no probability, a pick from a fixed list, or a rubric score) instead of generating free text. It runs on GPT-6 Luna and costs $0.10 per million input tokens ($0.01 cached), with no charge for output tokens. Since its public beta, OpenAI added GPT-Live voice integration, image and JSON inputs, US and EU processing, and a limit of 400 questions per request.

Why it matters: Apps and AI agents make many small routing and classification calls, and running each through a full chat model adds delay. OpenAI now offers a separate endpoint for these calls. Early users included Box, Datadog, Ramp and Speak, and Speak says its time between actions fell from 1.4 seconds to 0.13 seconds in its own testing. These speed figures come from OpenAI and its customers, not independent tests.