OpenAI makes decision API generally available with GPT-6 Luna – Unite.AI

OpenAI made its Decisions API generally available on October 22, 2026, the company announced, dropping the GPT-6-based endpoint Luna from public beta at a price of $0.10 per million input tokens. The API returns delimited, typed responses (a probability, selection, or rubric score) that applications use to classify content, route requests, and prioritize work.

OpenAI said the API focuses Luna’s intelligence on questions with limited answers, enabling near real-time decisions from text, images and structured inputs. The company said the endpoint provides such responses up to 10 times faster than those requested from GPT-6 Luna via the Responses API.

OpenAI introduced GPT-6 Luna alongside GPT-6 Sol on September 22, 2026, pricing Luna at $0.10 per million input tokens and $0.50 per million output tokens. The company said it has reduced API prices for both models by 50% compared to promotional GPT-5.6 prices and described the pair as an advance in the frontier of cost efficiency; continues to describe the GPT-6 Astra as its best model across the board. In the API, the two models are available as gpt-6-sol and gpt-6-luna.

How the Decisions API works

Requests go to POST /v1/decisions, and gpt-6-luna is currently the only supported model, according to OpenAI developer documentation. Each request pairs the shared input (a text string or an ordered series of messages, tool calls, and tool outputs) with between 1 and 400 questions, and the response contains an array of answers that echo the name the developer assigned to each question.

The API supports three types of questions. A predicate question returns a 0 to 1 probability that a condition is true, such as visible damage in a product photo. A choice question selects one of the developer’s predefined options and also returns the probabilities for each option and a confidence value. A scoring question evaluates an input against two to ten ordered rubric levels and returns the probability-weighted average of the level indices, so that a score can fall between the levels.

Developers can pass structured data directly through an input_json piece of content in a user message, and images can be served as embedded base64 data URLs or as publicly accessible HTTP or HTTPS URLs. Because conversation history can be provided as input, the API can evaluate a request in context along with custom tool and function calls and their outputs. The documentation states that file, audio, and item references are not supported, and the endpoint does not accept file_id input.

For decisions that depend on a previous response, the documentation instructs developers to submit separate requests. It also directs developers to structured outputs or function calls in the Responses API when they need a custom JSON object or tool call required by the model instead of one of the three typed responses.

Additions from the public beta

OpenAI said that since the public beta it has added native GPT-Live integration, out-of-the-box caching for reused context, and inference residency in the US and EU, as well as additional inference capability in several regions of Asia. Developers can now pass native JSON input, include images in question prompts, and ask up to 400 questions per request, double the previous limit.

The documentation’s list of new additions also includes hosted HTTP and HTTPS image input, conversation history evaluation with tool calls and output, organization usage reporting, and private link support. The usage endpoint, GET /v1/organization/usage/decisions, requires an admin API key with the api.usage.read permission and returns token usage and request counts that can be filtered or grouped by project, API key, or template.

With timely caching, repeated requests can reuse a shared input matching prefix at least 1,024 tokens long and receive discounted cache read prices; The content of the question, including question-specific images, is outside the reusable prefix. The API places an implicit cache breakpoint at the end of shared input by default, and developers can partition reuse with a caller-supplied prompt_cache_key or navigate to explicitly marked breakpoints.

With GPT-Live integration, a speech application can make decisions as soon as input or output transcripts arrive, and Live emits a decision event that the application can act on. OpenAI’s example configuration, added to session.triggers, selects an avatar’s facial expression from transcript updates, and the company depicts a customer support agent using the same model to decide whether to ask a clarifying question, route a request, or escalate an issue.

Reported use and initial evaluations

OpenAI said its own teams already apply the API across all product and business workflows: the dots team uses it to reduce response times and improve interactivity, the Personal Finance team uses it to classify transaction descriptions, and the Communications team uses it to classify coverage by theme and evaluate whether coverage reinforces or challenges key narratives, helping to identify what deserves a deeper review.

Customers mentioned in the ad reported speed and cost gains in their ratings. Box’s vice president of product management for platform and integrations, Biswaroop Palit, said the Decisions API combined with Box Extract enabled approximately 5x faster path selection on 30 certificate of insurance cases with the same accuracy as the Responses API, sending cases with less confidence to reviewers. AJ Stuyvenberg, staff engineer at Datadog, said the API risk-assessed agent-generated rshell commands before execution, delivering p95 latency 40 times faster than available frontier models with comparable accuracy. Ramp research engineer René Sultan said the API made attributing AI costs to underlying work 4 times faster and 30% cheaper than using GPT-6 Luna, without sacrificing accuracy.

Sierra engineer Ajeet Grewal said the API chooses the next action during parts of Sierra’s orchestration in live customer conversations at less than 280 milliseconds p95, about 3x faster and at a 90% lower cost per call than the company’s previous approach, with no decline in customer outcomes. CTO Andrew Hsu said integrating the Decisions API into GPT-Live reduced the time it took to perform the next action by more than 90% in the company’s tests, from 1.4 seconds to 0.13 seconds.

Pricing and data controls

The Decisions API costs $0.10 per million input tokens and $0.01 per million cached input tokens, and OpenAI only charges for input tokens, with no costs for output tokens. Regional processing premiums and input price multipliers apply for long contexts, and caching read and write rates match the standard gpt-6-luna prompt caching rates.

Regional processing is supported in the US and Europe (EEA plus Switzerland), data residency is available in all supported API regions, and zero data retention and HIPAA use are available to eligible customers. The API is accessible via the dedicated Decisions endpoint and a Decisions Playground on the OpenAI platform, where developers can test questions against their own text or images before writing code.