"System One model" is the name TypeSafe AI uses for AI models built for quick, bounded judgments: picking an option, scoring or answering yes or no, with probabilities, instead of writing text. The idea comes from Daniel Kahneman's System 1, fast and intuitive thinking. TypeSafe presents Jev as the first one.
A question from the app. Pick an answer.
Jev is an AI model made by TypeSafe AI.
What does Jev give you back?
That is what an LLM does. Jev returns typed values your program uses without interpreting prose.
For example, it picks an option from your list and tells you how sure it is. Your code decides what to do with it.
Jev does not generate code or free text. That is a job for an LLM.
In "Thinking, Fast and Slow", psychologist Daniel Kahneman describes two modes of thinking. System 1 is fast and intuitive: you can tell someone is angry just by looking at their face. System 2 is slow and deliberate: the one you use for a hard calculation or for reading the fine print of a contract.
TypeSafe calls System One the models that do System 1 work: quick, bounded judgments. Its model Jev takes a context and typed questions, and returns a choice, a score or a yes or no, with probabilities. It does not write, explain its reasoning or chat.
"The first System One model" is how TypeSafe presents Jev. It is the company naming its own category, so read it as the company's framing.
And it is no longer the only one. Laya, by Convai Innovations, is an open model with the same idea: it answers Choice, Score and Noul, accepts the same request format as Jev and can run on your own computer.
The name compares the model to a way of thinking; it does not describe how it is built inside. Being called System One tells you nothing about its layers or how it computes.
Nor does it mean every LLM is a "System 2". A chatbot can answer something simple right away, with no intermediate steps, and one that reasons step by step can still get it wrong. The label is about the kind of task, not about a brain.
TypeSafe sets Jev apart from a reasoning model, which generates intermediate steps before answering, even if they are not always shown. Jev returns a choice, a score or a yes or no, with probabilities, with no text or explanations.
It helps you pick the tool. If the task is classifying, scoring or choosing between fixed options, a model of this kind fits. If something has to be written, explained or created, that is a job for generative AI. And they can work together: one decides fast and, when it is unsure, the case goes to a person or another system.