OpenAI on Thursday made its Decisions API generally available, the company said, opening to all customers a service that returns fast, bounded answers that applications and agents can use to classify inputs, route requests and choose next steps.
The API was first announced at OpenAI’s September 29 developer day, and the general release adds voice and visual capabilities, TechCrunch reported, and OpenAI said developers can now include images in question instructions and connect the API natively to GPT-Live, its real-time voice product.
“The big unlock with Decisions API has been the speed, shrinking things from one to three seconds to 100, 120, 180 milliseconds,” Nikunj Handa, the product lead for OpenAI’s APIs, told TechCrunch.
Built on GPT-6 Luna
The API runs on GPT-6 Luna, currently the only supported model, through a dedicated endpoint, POST /v1/decisions, rather than the general purpose Responses API, according to OpenAI’s developer documentation. OpenAI said it can make decisions up to 10 times faster than GPT-6 Luna through the Responses API. It returns one of three answer types: a probability that a condition is true, a choice from a fixed set of options, or a score against a rubric.
Handa told TechCrunch the team did not build a specially calibrated decision model, but reconfigured Luna to speed up inference and lower cost.
OpenAI prices the API at $0.10 per million input tokens and $0.01 per million cached input tokens, with no charge for output tokens. Since the public beta, the company has added prompt caching, US and EU inference residency, native JSON inputs and a limit of 400 questions per request, up from 200, according to its announcement and changelog.
Customer results and rivals
OpenAI cited early users including Box, Datadog, Ramp, Sierra and Speak. Speak CTO Andrew Hsu said in OpenAI’s announcement that the GPT-Live integration cut the time to the next action from 1.4 seconds to 0.13 seconds.
The release puts OpenAI into a growing market for decision models, which TechCrunch described as fast classifiers that can replace larger language models in some workflows. TechCrunch said the category recently made TypeSafe, a startup founded by a former OpenAI researcher, a $7.6 billion company with its Jev models. Amazon Web Services earlier released Strands Decider 2B, an open source decision model for agents.
TechCrunch reported that OpenAI’s model trails several rivals on two independent decision model rankings, though Handa said Luna performs strongly on image classification.
Sources
- TechCrunch (Tim Fernholz): OpenAI makes its Jev competitor widely available as decision model competition heats up (October 22, 2026)
- OpenAI: Build more responsive apps with the Decisions API (October 22, 2026)
- OpenAI: Decisions developer guide (October 22, 2026)
- OpenAI: API changelog (October 22, 2026)
- OpenAI: API pricing (October 22, 2026)