Open· Last updated 2026-09-21

What is Jev?

Jev is TypeSafe AI’s flagship System One model: you send a state and typed questions; you get structured answers your code branches on. It is not a chatbot. This open-cluster page is the independent field-guide cut — complementary to the explainer at /explainers/what-is-jev, not a second official doc.

jev.pro is not affiliated with TypeSafe AI. Canonical: docs.typesafe.ai. We do not sell or issue API keys. TypeSafe’s launch post (15 September 2026) is a vendor source when we mention speed or “can’t hallucinate” (that phrase means out-of-schema, not factual truth).

Independent angle (cover, do not clone)

Win deltas (what-is-jev, system-one-concept, myth-faq, company-context): schema-first mental model + dual Py/TS + jaggedness caveats + GEO FAQ atoms under this open URL; short TypeSafe context without a brand-bait microsite. We do not mirror rival /what-is-jev/ IA.

Schema-first mental model

Think function call, not conversation:

1. Assemble a small state (string, JSON object, or array of text). No images, audio, or video.
2. Ask Choice, Score, and/or Noul in one questions map. They run in parallel on the same state.
3. Read choice / score / noul, the probabilities map, and (Choice/Score) confidence.
4. Your code acts, abstains, or calls an LLM to draft.

Official primer: a good question is a snap judgment a knowledgeable person could make in a few seconds. Decompose multi-factor work; combine in code.

English is the primary training language (models page). Other languages: test and watch confidence.

Short TypeSafe context (not a company microsite)

TypeSafe publishes Jev as the first System One model and documents HTTP POST /v1/systemone, GET /v1/models, official Python/JS SDKs, a Playground, and an agent skill. Founders-and-funding color belongs on their site. We only need: hosted model, not customer-LoRA’d, not trained on your requests per their Legal/data-handling pages — vendor claims; read the source.

Jaggedness in one paragraph

jev-1.13 is documented as strong on common-sense and weak on literal edge cases, math, dates, indirection, noisy state, and adversarial text. Keep arithmetic in code. Full list: jaggedness and TypeSafe.

Dual-stack first call (keys from TypeSafe, not us)

from typesafe_sdk import TypeSafeClient, Choice, Noul

with TypeSafeClient(model="jev-1.13.0") as client:
    result = client.system_one(
        state="User: I was charged twice. Please reverse the extra payment.",
        questions={
            "intent": Choice(
                instructions="Which intent fits?",
                criteria={
                    "refund": "Wants money back",
                    "status": "Order tracking only",
                    "other": "Neither",
                },
            ),
            "auto_ok": Noul(instructions="Is it safe to auto-reply with no human?"),
        },
    )
print(result.answers["intent"].choice, result.model, result.usage.input_tokens)

import { TypeSafeClient, choice, noul } from "@typesafe-ai/sdk";

const client = new TypeSafeClient({ model: "jev-1.13.0" });
const result = await client.systemOne({
  state: "User: I was charged twice. Please reverse the extra payment.",
  questions: {
    intent: choice({
      instructions: "Which intent fits?",
      criteria: { refund: "Wants money back", status: "Order tracking only", other: "Neither" },
    }),
    auto_ok: noul({ instructions: "Is it safe to auto-reply with no human?" }),
  },
});

List aliases with GET /v1/models. Public list price (vendor): $0.042 / million input tokens; output free — confirm on models.

What this page does not claim

• This page does not replace TypeSafe docs.
• No invented benchmarks.
• Open replicas are not Jev (OpenJev).

FAQ

Can Jev generate text? No. Official docs: no replies, code, or reasoning text. Use an LLM after the decision if you need prose.

Can Jev hallucinate? Outputs stay in your schema. That is not the same as being factually correct. Low-confidence or wrong labels still happen.

Is Jev an LLM? TypeSafe frames System One as a decision model, not a generator. We do not run a bake-off; read their AI primer.

Is Jev open source? The hosted model is not. Third-party interface clones exist and are not endorsed.

Can I fine-tune Jev? Does it train on my data? TypeSafe: no customer LoRA; not trained on customer requests/responses. Confirm Legal. Shape behavior with state and criteria.

Why is it called Jev? Naming lore belongs on TypeSafe’s materials. Operationally: System One model, typed questions.

Disclaimer

This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.

Open-cluster pages are independent field-guide notes. Replicas and third-party interfaces mentioned anywhere on jev.pro are not Jev and not endorsed. Hub: Open. Siblings: what is jev, how jev decides, why use jev. Canonical: https://docs.typesafe.ai.