OpenAI made its Decisions API generally available on Thursday, moving the GPT-6 Luna-powered endpoint out of public beta. The API is built for fast, bounded answers inside apps and agents, such as classifying an input, routing a request or picking a next action, and OpenAI says it runs up to 10x faster than posing the same question to GPT-6 Luna through the Responses API.
Pricing is input-only. OpenAI charges $0.10 per million input tokens and $0.01 per million cached input tokens, with no charge for output tokens. Requests go to a dedicated Decisions endpoint, and GPT-6 Luna is the only supported model.
Developers define a question and the answers it allows, then supply context as text, images or native JSON. The Decisions API returns one of three typed answers: a predicate, which is a probability from 0 to 1 that a condition holds; a choice among predefined options, with per-option probabilities and a confidence value; or a score, a probability-weighted average across an ordered rubric of 2 to 10 levels.
Key facts
- 01OpenAI made the Decisions API generally available on October 22, 2026, running on GPT-6 Luna.
- 02Pricing is $0.10 per million input tokens and $0.01 per million cached input tokens, with no output-token charges.
- 03OpenAI says the endpoint is up to 10x faster than asking GPT-6 Luna the same question through the Responses API.
- 04Each request can now carry up to 400 questions, double the limit during public beta.
- 05Speak cut its time to next action from 1.4 seconds to 0.13 seconds in GPT-Live testing.
The general-availability release adds several features since the beta. Each request can now carry up to 400 questions, double the previous limit. OpenAI also added prompt caching for reused context, inference residency in the US and EU, native JSON inputs, images inside question instructions and extra capacity in several Asian regions.
The most visible addition is native GPT-Live integration for voice apps. Decisions can run while a live transcript is still arriving, so an app can pick an avatar's facial reaction mid-conversation, or decide whether a support agent should ask a clarifying question, route the request or escalate.
OpenAI published latency results from early customers. Box reported roughly 5x faster route selection across 30 certificate-of-insurance cases at the same accuracy as the Responses API. Datadog saw 40x faster p95 latency with comparable accuracy when scoring the risk of agent-generated shell commands.
Ramp uses the API to attribute its AI costs. Sierra reported next-action selection under 280 ms at p95, about 3x faster and 90% cheaper per call than its previous approach. Speak cut its time to next action by more than 90%, from 1.4 seconds to 0.13 seconds, in GPT-Live testing.
“The Decisions API has made it 4x faster and 30% cheaper to attribute our AI costs”— René Sultan, Research Engineer, Ramp
OpenAI also runs the Decisions API internally. Its dots team uses it to cut response times, its Personal Finance team categorizes transaction descriptions, and its Communications team classifies media coverage by theme and by whether it reinforces or challenges key narratives.
The API is available now through the dedicated endpoint and a Decisions Playground. Regional processing is offered in the US and Europe, and Zero Data Retention is available to eligible customers. OpenAI introduced GPT-6 Luna alongside GPT-6 Sol on September 22, 2026.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
