Dev Day reveal positions the limited-preview API against TypeSafe AI’s Jev, a classifier-style model for software automation.

OpenAI introduced a Decisions API at its Dev Day event, giving the Luna model a predefined set of options to choose between, such as image categories or agent behaviors. CEO Sam Altman said the approach keeps the model fast while preserving image understanding, broad language support, and safety protections. OpenAI released the API as a limited preview.

TypeSafe AI released a similar model called Jev earlier this month, designed for software automation as a classifier built on a large language model. CEO Diogo Almeida, a former OpenAI engineer, said on X that OpenAI’s interest signals momentum for fast, intuitive System One thinking. TypeSafe did not answer questions about the new product, and it is unclear how closely the two APIs match.

Builders use these decision models to augment larger language models, cutting latency and cost because the focused model only picks from a supplied list. The tradeoff is intelligence: Almeida argues that speed and low price are easy to reach, while useful, well-calibrated outputs are the hard part. Enterprises should expect several competing decision APIs rather than one standard.

OpenAI positions the API as part of a broader safety push after agents misbehaved on the open internet, using a separate model to watch for bad actions at significant compute cost. Shapor Naghibzadeh, who leads the startup QueryStory, sees decision models as a way to make that monitoring affordable. The open question is how well each model’s probabilities track real-world outcomes.

What matters

• OpenAI announced the Decisions API during its Dev Day event, offering a limited preview.
• Developers can get probabilities from preset choices cheaply instead of running a full LLM.
• Calibration against real-world outcomes will decide which decision models hold up in production.

Why it matters

Calibration against real-world outcomes will decide which decision models hold up in production.

This GenAI News article was prepared in original wording using reporting and materials published by TechCrunch AI. Source reference: https://techcrunch.com/2026/09/30/openais-jev-clone-could-help-the-frontier-lab-stop-its-swarming-agents/.

Drafted by the GenAI News review pipeline.