Decision Models
A decision model takes some input and answers questions whose possible outputs are defined in advance. The result might be a choice, a probability for a yes-or-no question, or a score, rather than a free-form response.
Jev is TypeSafe AI's example. Its three question types are Choice, Score and Noul, a probability for a yes-or-no question (Almeida, 2026) (TypeSafeAI). TypeSafe says Jev is trained for calibrated decisions, but has not published enough about its architecture or training method to describe how it works internally (Almeida, 2026).
OpenAI's Decisions API offers a similar interface with choice, score and predicate questions, currently using GPT-6 Luna (OpenAI). A similar interface does not establish that the underlying models work the same way.
See Building an agent harness from scratch to pass Ian Goodfellow's intelligence test for a tool safety example, and How Jev Works for my explainer.
References
Diogo Almeida. Introducing system one models & jev. 2026. TypeSafe AI launch post, 15 September 2026. Accessed 11 October 2026. URL: https://typesafe.ai/blog/introducing-system-one-models-and-jev (visited on 2026-10-11). ↩ 1 2
OpenAI. Decisions. OpenAI API documentation, public beta. Accessed 10 October 2026. URL: https://developers.openai.com/api/docs/guides/decisions (visited on 2026-10-10). ↩
TypeSafe AI. Primitives (questions). Living documentation. Accessed 11 October 2026. URL: https://docs.typesafe.ai/primitives (visited on 2026-10-11). ↩