Jev
jev-latestUse Jev in a hosted Python Task to classify text, score it, or answer yes-or-no questions about it, with a confidence on every answer.
TypeSafe AI's System One decision model. It reads the prompt's text and answers typed questions about it. A pick-one question returns the chosen option with a confidence, a score question returns a value on your scale with a confidence, and a yes-or-no question returns the probability from 0 to 1 that the statement is true. Use it in prompts and tasks to classify or route input.
What Jev supports in qbash
- Typed answersAnswers pick-one, score and yes-or-no questions, each with a confidence.
What is qbash?
qbash is where you build Tasks: Python programs that run a business process on managed infrastructure. Your code runs the steps and calls a model like Jev only at the ones that need judgment, with integrations, memory and files available in the same script.
A developer writes the Task once, and anyone on the team runs it from a form, on a schedule, from a webhook or from chat. Each Task is versioned and every run is logged.
Specifications
- Context window
- 64Ktokens
Use Jev in a Python Task
Pass a question set as output_schema and Jev returns an answer for each question, with a confidence you can branch on in Python.
ticket = qbash.inputs.get("ticket")
decision = qbash.ai.raw(
ticket,
model="jev-latest",
output_schema={
"department": {
"type": "choice",
"instructions": "Which team should handle this ticket?",
"criteria": {
"billing": "Payments, invoices, or subscriptions",
"technical": "Bugs, outages, or integration problems",
},
},
"urgent": {
"type": "noul",
"instructions": "Does this ticket require an immediate response?",
},
},
)
route = decision["department"]
if route["confidence"] < 0.75:
qbash.output({"status": "needs_review", "decision": decision})
else:
qbash.output({"status": "routed", "department": route["choice"]})
Jev pricing
| Input | $0.042 per 1M tokens |
|---|---|
| Output | $0.00 per 1M tokens |
Standard rates in USD.
Automate Jev with qbash
Guides that use Jev
Jev vs OpenAI Decisions for entity resolution
A qbash Task matches company mentions from call transcripts to database records with Jev and OpenAI Decisions. Jev scored 93% and OpenAI Decisions 80% with the full catalog in every prompt. A Double Metaphone shortlist of up to 10 candidates brought both to 83% at about a ninth of the estimated cost.
Who the hell is Jev, and why is everyone talking about decisions?
Jev answers typed questions about text and returns probabilities, which makes it the newest version of the oldest AI in production, the classifier. How it compares with an LLM call that returns structured output, and what each costs at a million emails a month.
I built a Jev + Claude loop to stop AI slop
A qbash Task where Claude drafts a product launch email, Jev answers 12 review questions about it, and regex checks enforce the house style. Failed drafts go back for another round. The logs from two runs show where a decision model helped.
Meet Jev, the first non-LLM model on qbash
TypeSafe's Jev answers typed questions with a confidence score at $0.042 per million input tokens. This edition also covers Claude Opus 5.5, GPT-6 Sol and Luna, Sites, several triggers per Task, three new videos, and the AI Ticker for September 18 to 24.
Use Jev with qbash Tasks
A short walkthrough of TypeSafe's Jev decision model in qbash. Enable Jev on the Models page, create an inline Jev prompt with Pick one, Score, and Yes or no questions, and run it over four support tickets from a Python Task.
Why I'm paying attention to Jev
Jev answers bounded questions about text or structured data with typed answers and probabilities. In a qbash Task it fits beside code and LLMs, and we plan to measure it there before making claims about accuracy or cost.