OpenAI unveils Decisions API, a Jev-style model for fast, cheap AI decisions

OpenAI used its Dev Day event on September 29, 2026 to introduce a new product called the Decisions API. CEO Sam Altman mentioned it in passing and said the tool offers functionality similar to Jev, a model that TypeSafe AI released earlier in the month.

What the API does

Jev is a classifier built on top of an LLM: developers hand it a set of choices, and it returns probabilities cheaply and at high speed. Altman described OpenAI's product the same way, saying it gives the company's Luna model a predefined set of options to choose between, such as categories in which to classify an image or different 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.

TypeSafe's reaction

TypeSafe did not answer TechCrunch's questions about the new product. CEO Diogo Almeida, a former OpenAI engineer who co-invented reinforcement learning, joked on X about the start of the clone wars. He added that OpenAI's interest could be "a sign…that building in a System One compatible way is the future." System One is TypeSafe's term for fast, intuitive thinking; System 2 means deliberate reasoning.

Why cheap decisions matter

Today's LLMs are a poor fit for much software because they are comparatively slow and expensive; developers using Jev to augment them report faster, cheaper results.

Monitoring and securing AI agents is one likely application. After a series of incidents in which its agents misbehaved on the open internet, OpenAI added security measures, including a separate model to watch for bad actions at "significant compute cost." Shapor Naghibzadeh, who leads the startup QueryStory, thinks a model like Jev could do the job far more cheaply.

For a hackathon last weekend he built a demo that uses Jev to check each agentic action against the task it was given, blocking actions it was confident were bad, flagging others for review, and allowing the rest. In theory such monitoring could have stopped the Hugging Face incident. According to the report, it costs $2.94 with Jev versus $372 with a frontier LLM.

An open question

It is not yet clear how similar Decisions API is to Jev: OpenAI released it as a limited preview and developers have not been spotted testing it, though interest on X is evident. It is also not the only Jev-like tool on the market, as other startups roll out comparable models. A key question is how calibrated each model's outputs will be to real life.

Almeida says his company's moat is the synthetic data it creates to produce statistically useful outputs. "Fast and cheap is very easy, you know," he told TechCrunch last week. "If you want it really fast and cheap, use dice, right? Intelligence is the hard part, and my North Star is always pushing the intelligence-per-dollar Pareto curve."