OpenAI Launches Decisions API to Rival TypeSafe AI’s Jev Classification Model
Unveiled at Dev Day on Sept. 30, 2026, the API lets Luna choose from predefined options such as image categories or agent behaviors, targeting faster, cheaper software tasks. TypeSafe AI CEO Diogo Almeida called it “the beginning of the clone wars.”
OpenAI announced a new “Decisions API” at its Dev Day event on September 30, 2026, offering functionality similar to Jev, a fast and low-cost classification model released by startup TypeSafe AI earlier that month.
CEO Sam Altman described the API as a way to give OpenAI’s Luna model a predefined set of options to choose between — such as image categories or agent behaviors. “By focusing the model on that choice, we can make it extremely fast while keeping capabilities like image understanding, broad language support, and safety protections,” Altman said at the event.
TypeSafe AI’s Jev works similarly: developers supply a set of choices, and the model outputs probabilities cheaply and at high speed. TypeSafe CEO Diogo Almeida, a former OpenAI engineer, responded to the announcement on X by joking about “the beginning of the clone wars,” while suggesting OpenAI’s move could signal that “building in a System One compatible way is the future” — TypeSafe’s term for fast, intuitive decision-making as opposed to slower, deliberate reasoning.
The underlying case for these models is that standard LLMs are comparatively slow and expensive for many software tasks. Developers have been using Jev to augment LLMs, finding it faster and cheaper. Almeida told TechCrunch that speed and cost alone aren’t the differentiator: “Intelligence is the hard part, and my North Star is always pushing the intelligence-per-dollar Pareto curve.” He says TypeSafe’s moat is the synthetic data it uses to produce statistically useful outputs.
One emerging use case is monitoring AI agents. Following incidents where OpenAI’s agents misbehaved on the open internet, the company began using a separate model to watch for bad actions — at what it described as “significant compute cost.” Cybersecurity professional Shapor Naghibzadeh, who leads startup QueryStory, built a hackathon demo using Jev to screen agentic actions in real time, blocking high-confidence bad actions and flagging others. He found the approach cost $2.94 using Jev, compared to $372 with a frontier LLM.
OpenAI’s Decisions API is currently available as a limited preview, and direct comparisons with Jev remain limited. Other startups are also releasing similar models, and OpenAI is unlikely to be the last major tech company to enter the space.