Jev is a decision model from TypeSafe AI: instead of writing text, it answers typed questions with structured values and probabilities that software can use directly.
In plain words
Chat models produce text, so a program that wants a yes/no or a pick from a list has to parse that text and hope it is valid. Jev skips text entirely. You send it some state, such as a message or a record, plus questions of three kinds: pick an option, score against a rubric, or say whether a statement is true. It returns the answer with probabilities and a confidence value. TypeSafe calls this kind of model a System One model.
A quick example
Imagine a support system deciding where to send an incoming email. It asks Jev one choice question listing the teams. Jev returns the chosen team plus a probability for each option, and the code can route the email automatically when confidence is high and send it to a person when it is low.
Where you'll meet it
- Routing and triage inside software.
- Scoring or classifying records at scale.
Related
Sources
- DocsIntroducing System One Models & Jev, TypeSafe AI · read 29 Sept 2026
- DocsIntroduction, TypeSafe AI documentation · read 14 Oct 2026
Checked by WhatIsAI against the sources above · 14 Oct 2026. Spotted an error? Report it.