OpenAI announced on October 22, 2026 that its Decisions API is now generally available. According to OpenAI, the API is built with GPT-6 Luna and focuses the model on questions with bounded answers, which the company says lets applications make near real-time decisions from text, images, and structured inputs up to 10x faster than GPT-6 Luna through the Responses API. OpenAI frames the tool around the many small decisions applications and agents make in a workflow—classifying an input, routing a request, or choosing the next action—and states that their latency and cost add up at scale. Developers define a question and allowed answers, then supply context as text, images, or native JSON. The API returns one of three answer types: a probability (0 to 1), a choice from supplied values (with probabilities and confidence), or a score against a rubric. Since the public beta, OpenAI says it has added a native GPT-Live integration, prompt caching, inference residency in the US and EU, native JSON inputs, and support for up to 400 questions per request, double the prior limit. OpenAI cites internal use for transaction categorization and media-coverage classification. Rates and caching details are in OpenAI's pricing and availability guide.
- Generally available October 22, 2026, built on GPT-6 Luna via a dedicated Decisions endpoint
- Claims up to 10x faster than GPT-6 Luna through the Responses API
- Returns three answer types: probability, choice, or score
- Supports up to 400 questions per request and native GPT-Live integration
- Inference residency available in US and EU; prompt caching added since beta
What it means for you
This is a developer tool for making lots of small, repeatable yes/no or pick-one decisions inside software faster and cheaper—things like 'route this support ticket' or 'categorize this transaction.' If you run operations, it doesn't change anything you do directly; it changes how a technical team or vendor might build the automations behind your workflows. The claimed speed and cost gains only matter if you already have, or are building, an app that makes these kinds of calls at volume.
Try this
If a routing or classification step in your operation feels slow or inconsistent, map that step on paper first—what comes in, what decision gets made, who or what acts on it. A clear map tells you whether the problem is the tool making the call or an unclear process underneath it.
Who should care
Developers and technical teams building applications or agents that classify, route, or score inputs at scale; product owners who already have such apps in production.
Skip this if
You're not building software and have no in-house or contracted development—then this is background noise. The decisions that matter in your business (money, people, customers) should still have a person on them, not a probability score firing automatically.
Sources: OpenAI — read the original