Glossary Entry

System One Model

A model that reads an input and returns typed, calibrated probability distributions over predefined answers (a choice, a score, a yes-or-no) in a single parallel pass, instead of generating text.

Models LLMs Decision Making

Also called: System One models, System 1 model, System 1 models

Seed source: TypeSafe AI documentation

The term was coined by TypeSafe AI for its Jev model in September 2026, borrowing Daniel Kahneman’s name for fast, intuitive System 1 thinking as opposed to slow, deliberate System 2 reasoning. A System One model takes a state and a set of typed questions, answers every question in parallel against the same state, and returns numbers that code can branch on directly: a probability per option for a choice, a probability per level for a score, and a single probability for a yes-or-no question.

Because there is no decoding loop, latency does not grow with the length of the answer and outputs are guaranteed to match the requested schema. The trade is that such a model cannot generate text, explain itself, or carry out multi-step reasoning, so it is used alongside a language model for routing, gating, scoring, reranking, and judging rather than in place of it. Its value rests on the calibration of its probabilities, which should be measured on your own decisions before they are used to act unsupervised.