Jev 1.13 is the first public version of Jev, the System One decision model from TypeSafe AI. Instead of generating text, it returns typed answers — a yes/no with a probability, a choice from a defined set, or a score — with calibrated probabilities, for routing, classification and other decision points inside an application. Version 1.13 also reads images as an experimental input. It was released on 17 September 2026 and is available on AI/ML API as typesafe/jev-1.13.
typesafe/jev AliasAI/ML API serves Jev under two kinds of model id. typesafe/jev is a floating alias that points to the newest Jev release and currently resolves to Jev 1.14. typesafe/jev-1.13 is pinned: requests always go to version 1.13. Pin a version when decisions must stay reproducible — for example when thresholds were tuned against 1.13's probabilities or when an evaluation set must keep producing the same results — and move to 1.14 deliberately after re-checking those thresholds.
noul (yes/no with probability), choice, scorePOST /v1/decisionsA request carries a state — the text the decision is about — and a set of named questions, each with a type and plain-language instructions. Jev answers every question with a value of the requested type and a probability, so application code can branch on the result directly instead of parsing a sentence.
Only input is billed, at $0.0546 per 1M tokens. Jev does not generate text, so there are no output tokens to charge.
Routing: deciding which model, queue or handler should take a request, with a probability that can be used as a confidence threshold.
Classification inside product logic, such as flagging urgent support tickets, detecting intent or checking whether a message meets a policy.
Scoring, where an application needs a number on a defined scale rather than a written assessment.
Workloads that must stay reproducible: pinning typesafe/jev-1.13 keeps decisions and their probabilities stable while newer Jev versions ship under the typesafe/jev alias.
import requests r = requests.post( "https://api.aimlapi.com/v1/decisions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "typesafe/jev-1.13", "state": "Help! My payments have been failing for 3 days.", "questions": { "urgent": { "type": "noul", "instructions": "Is this an urgent support request?" } } }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/decisions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "typesafe/jev-1.13", "state": "Help! My payments have been failing for 3 days.", "questions": { "urgent": { "type": "noul", "instructions": "Is this an urgent support request?" } } }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/decisions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"typesafe/jev-1.13","state":"Help! My payments have been failing for 3 days.","questions":{"urgent":{"type":"noul","instructions":"Is this an urgent support request?"}}}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
Only input is billed, at $0.0546 / 1M tokens. Jev returns decisions rather than generated text, so output is free.
Yes. Jev 1.13 is live on AI/ML API as typesafe/jev-1.13, served through the POST /v1/decisions endpoint.
typesafe/jev-1.13 is pinned and always runs version 1.13. typesafe/jev is an alias for the newest Jev release and currently points to Jev 1.14, so its behaviour changes when TypeSafe ships a new version.
When decisions must stay reproducible: thresholds tuned on 1.13's probabilities, evaluation sets that must keep producing the same results, or audited workflows. Move to 1.14 after re-checking those thresholds.
Typed answers with calibrated probabilities: a yes/no (noul), a choice from a defined set, or a score.
Yes, as an experimental input. Jev 1.14 extends image support with Box questions, which ask where something is in an image.
Input is billed at $0.0546 / 1M tokens. Output is free, because Jev returns decisions rather than generated text.
