TypeSafe Jev: API, question types and use cases

Jev is the first System One model from TypeSafe AI, released in September 2026. You send a state (context) and a set of typed questions; it returns structured answers your code can consume directly. It is not a chatbot and generates no text — it only makes decisions.

This site is maintained by the independent community with no affiliation with TypeSafe AI. Key facts carry original sources and verification dates; unmeasured figures are labelled as official claims or illustrations, and community opinions are strictly separated from official claims (see About this site).

Understanding Jev

Identify the Jev model and find the official TypeSafe website, docs, and console. Jev overview and official links →

Three question types

Choice, Score, and Noul — and how to ask many questions in one request. Capabilities →

Guides

From curl to the official SDKs, with the full HTTP reference including errors, limits, and retries. API reference →

Use cases

Four original hands-on labs, each with synthetic samples, runnable code and recorded API results: model routing, content moderation, tool approval, ticket triage. Use cases →

Ecosystem

276 catalogued GitHub projects: official SDKs, community integrations, and open replicas (replicas ≠ open weights). Projects →

Video tutorials

21 videos in English and Chinese, titles and channels verified, sorted by topic. Tutorials →

Jev in three minutes

One request: state + questions

{
  "model": "jev-latest",
  "state": "The customer's ticket text…",
  "questions": {
    "department": { "type": "choice", "instructions": "Choose the primary support team", "criteria": { "billing": "…", "technical": "…" } },
    "urgent": { "type": "noul", "instructions": "Needs same-day handling" }
  }
}

Illustrative response shape (not a measured result)

{
  "answers": {
    "department": { "type": "choice", "choice": "technical", "confidence": 0.78 },
    "urgent": { "type": "noul", "noul": 0.91 }
  },
  "usage": { "input_tokens": 412, "output_tokens": 67 }
}

Key figures from the official Models documentation (verified 2026-10-11): input $0.042 per million tokens, output free; up to 64k tokens per request; the official launch blog claims 70–500 ms typical latency (historically checked on 2026-09-22); this is an official performance claim. See Capabilities and Limitations.

Where to start

• New to Jev: What is Jev → Capabilities → Get started
• Ready to code: API reference, JavaScript, Python
• Evaluating fit: Use cases and the FAQ