Jev AI · TypeSafe System One model

Jev AI

Try the Jev model: probabilities, not prose.

Jev AI runs Jev, the TypeSafe decision model, on your own text. Ask yes/no, choice and score questions and get calibrated answers your code can act on. Then call the Jev AI API with the same request.

Try a Jev AI demo Jev AI API

3Question types: yes/no, choice, score

~0.5 sTypical Jev answer time

5 freePlayground runs to start

Jev AI playground

Run Jev on a real case

Recorded answers are real Jev output. Edit anything and run it live: 1 credit per run, 8 questions max.

Route a customer message to a team, flag urgency and read frustration in one call.

State the text Jev reads · plain text or JSON235 / 8,000

Add question

Jev answersRecorded 2026-09-18 · jev-1.13.0 · 1.50 s · 496 input tokens

Choiceteam confidence 100%

Which team should handle this ticket?

billing

100%

technical

sales

account

Yes / Nourgent

Does the customer need a response today?

96% yes

Scorefrustration confidence 100%

How frustrated is the customer?

3.00Angry and ready to churn

0 · Calm

1 · Mildly annoyed

2 · Frustrated

3 · Angry and ready to churn

100%

Yes / Nochurn_risk

Does the customer threaten to leave or switch providers?

98% yes

What is Jev AI?

Jev AI answers typed questions about your text.

Jev is TypeSafe's first System One model. Instead of writing an answer, Jev reads your text (the state), evaluates every typed question you ask in parallel, and returns structured answers: a yes/no probability, a choice with a full probability distribution, or a score on your own rubric, each with a confidence value. Jev AI is a hands-on playground for those decisions plus a metered, TypeSafe-compatible Jev API.

Read: what is Jev?

01: Typed questions, typed answers
Ask Jev a yes/no (noul), choice or score question and get a number or label your code can branch on, not a paragraph to parse.
02: Calibrated probabilities
Every choice and score comes with the full probability distribution and a confidence value, so you can route low-confidence cases to a human.
03: Many questions, one call
Jev reads the state once and answers every question in parallel. Triage, moderation and scoring fit in a single request.
04: Same request as the TypeSafe API
The Jev AI endpoint accepts the TypeSafe System One request shape, so playground experiments move straight into code.

Three question types

Every Jev AI answer is typed.

Yes / No (noul)

Jev AI returns the probability that the answer is yes, from 0 to 1. Threshold it in code.

"urgent": { "type": "noul", "noul": 0.96 }

Choice

Jev AI picks one option from the set you define and returns every option's probability plus confidence.

"team": { "choice": "billing",
  "probabilities": { "billing": 0.99, … },
  "confidence": 0.98 }

Score

Jev AI rates the text on your ordered rubric. The score is probability-weighted, so it can land between levels.

"frustration": { "score": 3.0,
  "legend": { "0": "Calm", … },
  "confidence": 0.93 }

How it works

From a Jev AI demo to production in four steps.

Jev AI is built for decisions your code can act on directly: routing, moderation, triage, scoring and guardrails, with probabilities instead of prose.

01
Pick a Jev demo or start blank
Load a real case such as ticket triage, moderation or lead scoring, or paste your own text.
02
Shape the questions
Edit the yes/no, choice and score questions, their options and rubric levels.
03
Run Jev and read the numbers
See the selected option, every probability and the confidence for each answer in about half a second.
04
Ship it with the Jev API
Copy the generated request, create a key and call the same Jev model from your backend.

Jev AI vs a chat LLM

When to reach for Jev AI.

	Jev AI (System One)	Chat LLM
Output	Probability, choice or score with confidence	Free text you must parse
Many decisions	All questions answered in parallel in one call	One prompt per task, or a long structured prompt
Uncertainty	Calibrated distribution for routing and review	No calibrated probability by default
Speed	About 0.5 s in our recorded Jev AI demos	Grows with the length of the generated answer
Best for	Routing, moderation, triage, scoring, guardrails	Writing, open-ended reasoning, conversation

Pricing

Start free, then pay per Jev AI run.

New accounts get 5 free runs. A playground run is 1 credit; the API is 1 credit per 1 million input tokens, and output tokens are free.

See Jev AI plans

Output: Probabilities and labels
Jev returns typed answers with confidence rather than generated text.
Playground: 1 credit per run
Up to 8 questions and 8,000 characters of state per run.
Jev API: 1 credit per 1M input tokens
Metered on the input tokens Jev reports; output tokens are free.

FAQ

Questions about Jev AI

Jev AI is an independent service and is not affiliated with or endorsed by TypeSafe. Jev answers are probabilistic; review important decisions before acting on them.

What is Jev AI?

Jev AI (jev-ai.pro) is an independent playground and API for Jev, the System One model from TypeSafe. You give Jev a piece of text and typed questions; it returns yes/no probabilities, choices and scores with calibrated confidence.

What is the Jev model?

Jev is TypeSafe's flagship model and the first System One model. It is trained for fast structured decisions instead of text generation: it evaluates many narrow questions about one state in parallel and reports how sure it is.

Is Jev AI affiliated with TypeSafe?

No. Jev AI is an independent service that calls the official TypeSafe API. Jev and TypeSafe are names of their respective owner.

How does the Jev API on this site work?

Create an API key on the Jev API page and send the same JSON you would send to TypeSafe (state, model, questions) to https://jev-ai.pro/api/v1/systemone. Usage is billed from your Jev AI credits by input tokens.

How much does Jev AI cost?

New accounts get 5 free playground runs. After that each playground run costs 1 credit and the API costs 1 credit per million input tokens. Credits come from the monthly plans or one-time packs on the pricing page.

Can Jev read images or PDFs?

No. Jev accepts text only: a string, a JSON object or an array. Convert other content to text or structured fields before sending it.

Which languages does Jev support?

English is Jev's primary language and where it is most accurate. Other languages work but less reliably, so test on your own data and watch the confidence value.