Four questions, one charge
10 credits for one to four questions; each one past four adds 2. The Run button shows the price before it runs.
Every paid plan runs all three answer types, works in the playground and over the API with your own keys, and never charges for a request that fails.
No hidden fees • Cancel anytime • Failed runs refunded
Pay safely and securely with
Payments are processed by Stripe over an encrypted checkout. Your card details are handled by Stripe and are never stored by Jevx.
Charges appear as “JEVX” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for. Operator details, the engine behind the answers, and the refund window are on the about page.
Powered by
One balance covers choice, score and noul alike, and up to four questions share one charge.
What you can ask for
Where you can run it
Jev AI pricing guide
Up to four questions ride on one charge; each one after that adds two credits. Failed requests cost nothing.
| Scope | What you send | Credit cost |
|---|---|---|
| Request | One state, one to four questions | 10 credits — one decision. |
| Request | One state, five to 16 questions | 10 credits plus 2 for each question past four — 16 questions is 34 credits. |
| Free | A new account | 20 credits — 2 requests, every answer type, no card. |
| Free | A 7-day check-in cycle | 110 credits — 11 requests, claims climbing 1·1·2·1·1·2·3. |
| Refund | A failed request | 0 credits. |
choice, score and noul, all unlocked at $0 and no card. Jev AI pricing charges for how many decisions you run — never for a shape of answer.
Jev AI pricing opens with two decisions waiting — enough to paste your own text in, read the probabilities and copy the request. Check in the next day and the claim climbs: 1, 1, 2, 1, 1, 2, 3. Check-in decisions are spent before anything you paid for and expire 30 days after you claim them.
Create an accountNot a market rate we matched — Jev AI pricing is built up from one measured call on 2026-09-21.
464 input tokens billed, 72 output tokens free. The input is the whole bill, which is why up to four questions can share one state and one charge in Jev AI pricing.
Run one nowEvery number came off a live request, not a launch post. Jev AI pricing shows the charge before a credit moves.
10 credits for one to four questions; each one past four adds 2. The Run button shows the price before it runs.
Our measured round trip on a short ticket with three typed questions.
Jev AI pricing counts input only. The answer, the probabilities and the confidence cost nothing.
Nothing is deducted for a request the model never answered. Built in, not a courtesy.
The cheap way to ask
The questions share the state, and the state is most of the bill. Sending them separately pays for the same text again and again.
{
"state": ticket,
"questions": {
"sentiment": { "type": "choice", … },
"urgency": { "type": "score", … },
"needs_human": { "type": "noul", … },
"category": { "type": "choice", … }
}
}{ "state": ticket, "questions": { "sentiment": … } }
{ "state": ticket, "questions": { "urgency": … } }
{ "state": ticket, "questions": { "needs_human": … } }
{ "state": ticket, "questions": { "category": … } }
// same answers, four round trips,
// and the ticket text paid for four timesUp to four questions share the 10 credits; a request can carry sixteen, at 2 credits for each one past four.
What people ask before they put a card in.
10 credits each, so a thousand decisions is 10,000 credits — about 0.7% of a Creator month, or $0.13 at the Creator rate. Each of those decisions can carry up to four questions for that same 10.
Yes, behind an account and with no card: 2 decisions the moment you sign up, then a daily check-in that climbs across a seven-day cycle to 11 more. Check-in decisions are always spent before anything you paid for.
Because those need fine-tuning to reach usable accuracy on your labels, and a fine-tune is a dataset, a training run and a redeploy every time your categories move. Jev is zero-shot: the criteria live in the request, and one costs a fraction of a cent.
One developer reported $0.40 for 11.2M uncached input tokens on a real classification job. Our own measured call was 464 input tokens at $0.0000195, with output tokens free.
A frontier chat model bills you for the paragraph it writes back. Jev bills input only — output tokens are free — so a judgement that used to cost a full generation now costs the text you sent and nothing else.
No. Jevx is an independent tool built on the Jev API and is not affiliated with TypeSafe AI.
Create an account, run two on your own text, no card. Check in for more while you decide. Pick a plan when you need volume; a failed request is never charged.