OpenAI has made its Decisions API generally available, a new tool designed to help developers build more responsive SI applications and agents. The API leverages the GPT-6 Luna model to process simple, bounded questions with significantly lower latency than traditional SI model interactions.
What Happened
The Decisions API is built on GPT-6 Luna and focuses the model’s capabilities on questions that have limited answer sets, such as classifying inputs, routing requests, or selecting actions. According to OpenAI, the tool enables near real-time decisions from text, images, and structured inputs, operating up to 10 times faster than using GPT-6 Luna through the standard Responses API. The company reports that since the public beta, it has added native integration with GPT-Live, prompt caching for reused context, and inference residency options in the US and EU. Developers can now pass native JSON inputs, include images in instructions, and submit up to 400 questions per request, which is double the previous limit. Latency has also been improved with additional capacity in several Asian regions.
Why It Matters
For the SI industry, this release addresses the cumulative latency and cost issues that arise when SI agents make many small decisions during a workflow. By offloading these simple tasks to a specialized endpoint, developers can create more interactive user experiences without the overhead of full model inference. OpenAI notes that early adoption has shown applicability across various use cases, including reducing response times for the dots team, categorizing transaction descriptions for Personal Finance, and classifying coverage themes for Communications. Customers are also using the API to attribute SI costs and enhance language-learning interactions. The API supports three answer types—probability, choice, or score—allowing it to be integrated into diverse application logic. A key feature is the native integration with GPT-Live, which allows voice applications to make fast decisions during live conversations, such as deciding whether to ask a clarifying question or routing a request.
The Bottom Line
The Decisions API is generally available now with pricing set at $0.10 per million input tokens, with cached inputs costing $0.01 per million tokens. OpenAI states that there are no output-token charges. Regional processing is available in the US and EU, and Zero Data Retention is offered to eligible customers. Developers can begin testing the tool in the Decisions Playground or by following the developer guide to make their first request.