OpenAI introduces Decisions API for faster choices in automated work
By Automate Basics, written with AI from TechCrunch's original
OpenAI has introduced a limited preview of Decisions API, which lets its Luna model choose from options set in advance. For teams building automation, the proposed uses include classifying images and selecting an AI agent’s next behavior.
What changed with OpenAI’s Decisions API?
OpenAI introduced Decisions API as a way for its Luna model to choose among options supplied in advance, TechCrunch reports. OpenAI gave image categories and possible AI agent behaviors as examples. Rather than asking a model for an open-ended response, a developer could give it a defined set of choices.
According to TechCrunch, OpenAI says narrowing the task is intended to make the model fast while retaining image understanding, language support and safety protections. That makes the announcement relevant to automated work that repeatedly needs a choice from a known list. TechCrunch compares the idea with TypeSafe AI’s Jev, but says it is not yet clear how closely the products work alike.
Who can use Decisions API, and what does it cost?
OpenAI has released Decisions API as a limited preview, according to TechCrunch. The report does not say who can enter that preview, which plans include the API or what OpenAI charges for it. Teams should not assume it is available in an account they already use.
The stated examples point toward people who build or manage automated workflows, rather than someone looking for a new chat feature. A workplace team might care if it needs to sort incoming images or keep an AI agent’s possible actions within a defined set. Those are potential uses based on OpenAI’s examples, not reported deployments or evidence that the API performs reliably in those settings.
How do I try Decisions API at work?
A team with access to the limited preview could start with a low-risk classification task, such as assigning incoming product photos to categories it has already defined. That follows OpenAI’s image-classification example without depending on the model to invent a new category or make a consequential decision.
Before relying on the output, compare its choices with examples that staff have already checked. Look closely at ambiguous images and cases where none of the available categories fits. If the intended use involves choosing an AI agent’s behavior, keep a person responsible for reviewing actions that could have harmful effects. TechCrunch says it has not yet seen developers put Decisions API through its paces, so the report does not establish how accurately it makes choices in workplace tasks.
Frequently asked questions
What does OpenAI’s Decisions API do?
According to TechCrunch, OpenAI’s Decisions API lets its Luna model select from options a developer provides in advance. OpenAI mentioned classifying images and choosing among possible AI agent behaviors. The report does not establish how well those choices work in practice.
Can my team use Decisions API now?
TechCrunch reports that OpenAI released Decisions API as a limited preview. The report does not identify eligible users, included plans or a price. A team interested in testing it would first need to confirm that it has access.
Is Decisions API the same as TypeSafe AI’s Jev?
TechCrunch compares Decisions API with Jev because both concern fast choices from predefined options. The report says the products’ similarity is still unclear and that it has not seen developers test OpenAI’s preview. Teams should not assume the two have the same outputs or performance.
Related reading
OpenAI adds virtual try-on and saved products to ChatGPTTechCrunch reports that OpenAI has launched virtual try-on and product Favorites in ChatGPT worldwide. People comparing clothing and accessories, including retail staff assembling looks, can preview items using their photos and save products for later.
Google adds spoken camera guidance to Gemini Live on AndroidGoogle’s Guided Vision is launching in Gemini Live on compatible Android devices. It describes what a phone camera sees aloud, helping people with low vision read small text, locate objects and ask follow-up questions.
Google makes Meet and Teams meeting room devices interoperableGoogle has made meeting access between Google Meet hardware and Microsoft Teams Rooms devices generally available on compatible Android-based equipment. Workplaces can use those room devices to join meetings hosted on the other service.
Written with AI from TechCrunch's original and published after automatic checks: every figure here appears in the original, and no sentence is copied from it. The picture is AI-generated. The original is the authority.
Source: OpenAI’s Jev clone could help the frontier lab stop its swarming agents, TechCrunch, 30 Sept 2026.