Decisions API

A GPT-6 Luna-powered endpoint that picks among a developer-defined set of answers in ~150ms instead of generating free text.

OpenAI · Agent frameworks and orchestration · framework

Published in Feed

Overview

OpenAI announced the Decisions API in limited preview at its September 29, 2026 DevDay, built on a specialized version of GPT-6 Luna focused on bounded, real-time decision-making. Rather than generating free text, the endpoint takes a context (text or image) plus a closed list of developer-defined answer options and returns the most likely option with a confidence score, for classifying content, routing requests, or choosing an agent's next action. OpenAI reports the specialized Luna variant returns a classified answer in roughly 150 milliseconds, versus about 1.6 seconds for a standard Luna call. On October 8, 2026, OpenAI opened the Decisions API in public beta on GPT-6 Luna, with pricing of $0.10 per million input tokens and no charge for output tokens. On October 22, 2026, OpenAI moved it to general availability with GPT-Live voice integration, prompt caching, JSON and image inputs, up to 400 questions per request, US and EU regional processing, and Zero Data Retention for eligible customers.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Closed-answer-set decision endpoint instead of free-text generation
  • Accepts text or image context plus a developer-defined list of allowed answers
  • Returns the selected option with a confidence score
  • ~150ms response time versus ~1.6s for a standard GPT-6 Luna call
  • Generally available since October 22, 2026: up to 400 questions per request, JSON and image inputs, prompt caching, GPT-Live voice integration
  • US and EU regional processing and Zero Data Retention for eligible customers
  • Built for real-time content classification, request routing, and agent next-action selection

Use cases

  • Low-latency content classification inside an application pipeline
  • Routing user requests to the correct handler or agent
  • Choosing an autonomous agent's next action from a constrained option set

Pricing

GA: input tokens only, $0.10 per million ($0.01 cached) for gpt-6-luna; output tokens not charged.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
29 Sept 2026
Record updated
23 Oct 2026
Record last checked
23 Oct 2026

Explore the Feed

Other entries filed under agent frameworks and orchestration; inclusion does not imply equivalent features.

  • Workday Agent Harness

    Workday's generally available environment for building enterprise agents with isolated runtimes, memory and governed HR/finance data access.

  • Agent Bricks

    Databricks' no-code system for building and auto-optimizing domain-specific AI agents, now generally available with synthetic eval-data generation and multi-agent orchestration.

  • Muse Agents SDK

    Meta's open-source SDK for building and orchestrating multi-agent systems on top of its Muse model family.

Browse the Feed