OpenAI Decisions API now available on AI Gateway

OpenAI’s Decisions API is now available on AI Gateway through an OpenAI-compatible /v1/decisions endpoint, allowing users to integrate decision models alongside existing language models. The API supports models like GPT-6 Luna Decisions, which return probabilities, choices, and scores instead of generated text, making it suitable for routing, triage, guardrails, and rubric scoring. Users can submit predicate, choice, and score questions against a shared input in a single request, streamlining decision-making workflows.

The Decisions API operates under a separate model ID, openai/gpt-6-luna-decisions, distinct from the text-generation model openai/gpt-6-luna. Requests to the decisions ID are billed at the Decisions API’s rates, ensuring transparent pricing for structured decision outputs. Developers can call the API using the OpenAI SDK, AI SDK, HTTP API at /v1/evaluate, or the AI CLI, with specific version requirements for each tool.

Integration with the OpenAI SDK requires version 7.30.0 or later for JavaScript, or 3.26.0 or later for Python, while the AI SDK needs experimental_decide with ai 7.0.128 or later. The API functions similarly to OpenAI’s Decisions API, with questions keyed by name and boolean types for yes/no queries. Answers are returned as an array in question order, with each entry matching the corresponding question’s name.

Users can select from multiple decision models, including Jev, Liquid d1, and Laya, by browsing AI Gateway or consulting decision model guides. The OpenAI-compatible Decisions API reference provides details on request limits and error formats, while the AI SDK documentation outlines experimental_decide usage. This expansion broadens AI Gateway’s capabilities for structured decision-making across applications.