Jev AI
Jev: TypeSafe's decision model, explained
Jev is a decision model built by TypeSafe. Instead of generating text like a chat model, Jev takes some state (text or JSON) plus one or more typed questions, and returns typed answers with calibrated probabilities. TypeSafe describes it as a "System One" model: fast, intuitive judgments rather than slow step-by-step reasoning.
That makes Jev a good fit for the many small decisions inside software: which queue a ticket belongs to, whether a lead is qualified, whether a message is spam, or whether an AI agent's proposed action needs a human to approve it.
Get Early Access to ClassifierHub: 2× credits in your first paid month.
How Jev works
You send Jev a state and a map of questions. Each question has instructions (what to judge) and criteria (what each possible answer means). Jev answers every question in a single request and returns a probability distribution over the answers you defined, never free text.
Because the output is typed, there's nothing to parse and no JSON prompt to maintain. Your code reads a label or a probability and decides what to do next.
The three question types
Jev supports three primitives, and most real-world decisions can be expressed with them:
- Choice: pick one option from a set you define, such as billing, technical or other. Returns the winning option, a probability for each option and a confidence value.
- Noul: a yes/no proposition such as "is the customer asking for a refund?". Returns the probability that the answer is yes.
- Score: place the input on an ordered scale of 2 to 10 levels you describe, such as severity or lead fit. Returns a probability-weighted position plus per-level probabilities.
What Jev is good at, and what it isn't
Jev is designed for high-frequency operational decisions: routing, classification, moderation, gating and workflow branching. It's fast (OpenRouter reports responses in the hundreds of milliseconds) and inexpensive because output tokens are free.
It is deliberately not a chat model. Jev doesn't write text, explain its reasoning, call tools or extract fields. Arithmetic, date comparisons and threshold logic belong in your code. The best pattern is: code prepares clean state, Jev judges, code acts on the probabilities.
Using Jev through ClassifierHub
ClassifierHub is a decision layer built on Jev. It adds ready-made classifier templates, saved and versioned decisions, API keys, credits, batch processing, MCP tools for AI agents and integrations such as Excel, so you don't have to build that plumbing around the raw model.
curl https://classifierhub.com/v1/classify \
-H "Authorization: Bearer $CLASSIFIERHUB_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": "Hi, I was charged twice for my subscription this month.",
"instructions": "Which team should handle this ticket?",
"options": ["billing", "technical", "other"]
}'Frequently asked questions
Related guides
Last updated 2026-09-25. ClassifierHub is an independent product built on top of the Jev decision model, accessed through OpenRouter. It is not affiliated with or endorsed by TypeSafe or OpenRouter.