Jev

Give it a sentence and a fixed set of answers, and it settles on one — instantly, silently, without a syllable of explanation.

Jev never argues its case. It doesn't write, reason aloud, or hedge in prose. It reads the state you give it and returns a typed answer: a chosen option, a place on a scale, or the odds of a single yes. Three shapes, tried below — each with the option to run the identical question through an ordinary LLM, so you can see the cost side by side.

Choicetype: choice

Picks exactly one option from a named set of up to 255, and returns a probability for every option it didn't pick, along with a confidence score for the one it did.

Scoretype: score

Places the state on an ordered rubric you describe in words, from two levels up to ten. The result can land between two levels rather than being forced onto one.

Noultype: boolean

A single yes-or-no question. Returns one probability that the statement is true — no forced verdict, just how likely.

Every run below calls Jev — and, where you ask for one, a comparison model — live through the AI Gateway. Nothing is cached or scripted. The comparison model reports its own probabilities; unlike Jev's, they are not calibrated.