Frequently asked questions
The questions that come up most about Jev and JevStation, answered directly. Everything here is also covered in more depth in the docs.
What is Jev?
Jev is the System One model from TypeSafe. It is not a chat model: instead of generating text, it takes a state — a support ticket, a review, a spec or a JSON record — plus typed questions, and returns a typed answer for each question along with the probability distribution behind it.
How is Jev different from a text-generating LLM?
A text-generating LLM produces prose that you then have to coerce into structured output and parse back. Jev returns the judgement itself as a typed value, so there is no parsing step and no malformed output to handle. It also returns a calibrated confidence value on every Choice and Score answer, which a prose model only gives you if you ask for it in the prompt and then trust the number.
What are Choice, Score and Noul questions?
They are the three question types Jev answers. A Choice question picks one option from a list you define. A Score question rates the state against an ordered rubric. A Noul question returns the probability that a statement is true, from 0 to 1.
How many questions can one evaluation ask?
Up to 12. Every question is evaluated in parallel against the same state, so asking twelve costs barely more time than asking one.
What does the confidence value mean?
Confidence collapses the shape of the probability distribution into a single number from 0 to 1. A distribution concentrated on one outcome is confident; one spread across several options is not. You can threshold on it directly: act automatically on a high value, ask for confirmation in the middle, and route to a human when it is low.
How much does one evaluation cost?
A standard evaluation costs 1 credit — state up to 8,000 characters and up to 5 questions. A larger evaluation costs 3 credits — up to 24,000 characters and up to 12 questions. New accounts receive 200 free credits, which is enough for 200 standard evaluations.
Is there a subscription?
No. Credit packs are one-time purchases: $9.90 for 5,000 credits, or $15.90 for 12,000. There is no recurring charge, no monthly commitment, and nothing to cancel. Purchased credits stay valid for 12 months.
Do I need my own model API key?
No. JevStation runs Jev for you, so creating an account is enough to start evaluating in the playground — there is no provider key to obtain and no local setup. If you would rather call the model directly, without JevStation metering, TypeSafe's own API is the supported path.
What happens if an evaluation fails?
Credits are consumed when the answers land. If the provider errors partway through, the run is refunded automatically and costs 0 credits.
What is Jev bad at?
Jev cannot write. It will not draft a reply, generate code, summarise a document in prose, or hold a conversation — those are jobs for a generative model. It is also a poor fit when you cannot enumerate the answer space in advance, because if you cannot list the possible options there is no Choice question to ask. Use it for judgements your code will branch on.