What is Jev?
Jev is TypeSafe AI’s first public System One model. You send a state (the material to judge) and a map of typed questions. Jev returns one typed answer per question, with probabilities your code can branch on. It does not write replies, produce code, or generate explanations of its reasoning.
This page is the definition-first briefing for builders who already heard the name and need a production-minded contract. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Canonical facts live on docs.typesafe.ai. We do not sell or issue API keys.
Definition
TypeSafe’s launch post (15 September 2026) frames Jev as “a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out.” The same post says Jev “gives up string generation,” is optimized for structured outputs, and “can’t hallucinate” in the sense that outputs are constrained to the schema you supplied.
Read that slogan as schema-constrained, not factually correct. A Choice can still pick the wrong department. It cannot pick a department that is not in criteria. Schema-safe ≠ correct. Always keep that distinction when someone pastes “can’t hallucinate” into a design doc.
Mental model: a function call, not a conversation
The useful contract is closer to a function than a chat thread. You declare the answer space up front with three primitives — Choice, Score, and Noul — evaluated independently and in parallel against one state. Code owns control flow. System One is the class name (after Kahneman’s fast judgments); Jev is the first public instance.
English is the primary training language. Input is text only. Current versioned id: jev-1.13.0 (aliases jev-latest and jev-preview pointed there on the public models page we used).
What you send
Every evaluation request has three top-level fields:
model— usuallyjev-latest, or a pinned version such asjev-1.13.0.state— a string, JSON object, or array of text. Images, audio, and video are not accepted.questions— a map of IDs you choose to Choice, Score, or Noul questions.
Question IDs are for your code. TypeSafe’s primitives page says they are not sent to the underlying model. Write the complete question in instructions.
What you get back
| Question | Typed fields |
|---|---|
| Choice | choice, probabilities, confidence |
| Score | score, legend, probabilities, confidence |
| Noul | noul (0–1). No separate confidence field. |
All questions in one request see the same state and are evaluated independently and in parallel. Adding questions costs the extra question tokens; docs say it barely changes response time (vendor claim).
What Jev is not
- Not a drop-in ChatGPT replacement.
- Not a generator for emails, code, or JSON documents.
- Not a calculator. The jaggedness note for
jev-1.13tells you to keep arithmetic in code. - Not an official product of this site. jev.pro is independent and unofficial.
TypeSafe publishes vendor figures you should label as such: $0.042 per million input tokens, output free; 70ms–500ms typical end-to-end; “40x–200x” on System One shaped queries; 64k / 32k context. This site does not add independent benches.
Start at docs.typesafe.ai, then the quick start. First call notes: Your first Jev API call. If we disagree with the docs, the docs win.
FAQ
Does Jev generate text if I ask nicely in instructions?
No. Official docs are explicit: System One models do not write replies, produce code, or generate reasoning text. Chaining Choices to spell a sentence is documented as slow and weak.
What does “can’t hallucinate” mean here?
Outputs stay inside the schema you supplied. It does not mean the chosen option or Noul is factually right. Schema-safe ≠ correct.
Is Jev the same as System One?
System One is the class (fast, structured decisions). Jev is TypeSafe’s first public model in that class. See System One.
Which model id should I send?
Pin jev-1.13.0 if you have tuned thresholds. Aliases move when TypeSafe ships a release. Log the response model field.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.