Developers automating routing or approvals may find it far cheaper than chat models, but it's early access from a waitlist.

Kenning's take

What it does

Jev is the first System One model from TypeSafe AI, built to make fast, structured decisions that software can act on directly instead of generating text. Each decision comes with a confidence estimate so a program can act when confidence is high and escalate when it is not. TypeSafe says it is two orders of magnitude faster and cheaper than chat models on these tasks; it is available through the TypeSafe API in early access.

  • Typed, structured decisions instead of free text
  • Confidence estimate with every decision
  • Trained with Reinforcement Learning for Calibrated Decisions
  • 64K token context per request
  • Priced per billion input tokens

Who it suits. Developers automating repeated decisions in software, such as routing, classification and approvals.

Key facts

Category
AI models
Made by
TypeSafe AI
How to get it
early access through the TypeSafe API console, with developers admitted from a waitlist
Works on
API
API price
$42 per billion input tokens ($0.042 per million) for jev-1.13.0

FAQ

What is Jev?

A model from TypeSafe AI that outputs typed decisions with confidence scores for software to use, rather than text.

How much does Jev cost?

Jev 1.13 is $42 per billion input tokens, or $0.042 per million.

How do I get access to Jev?

Through TypeSafe's API console; it is in early access with a waitlist.

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.