Is this an independent test or a gateway's write-up?

A gateway write-up. Requesty resells access to the model, the post restates TypeSafe's documentation and embeds the founder's own announcement. Its comparison table is explicitly described as a comparison of roles, not an accuracy ranking.

How is this different from just asking an LLM for JSON?

The post concedes LLMs can already return schema-constrained JSON. The claimed difference is the training objective: RLCD targets decisions and useful probability estimates rather than producing a preferred conversational response.

For a severity scale, do I use Choice, Score or Noul?

Score, for an ordered rubric. The post's guidance is that the level descriptions are what make a scale meaningful, since a bare one-to-ten leaves far more to interpretation, and probabilities come back across the levels.

What does this actually cost at real volume?

The entry copy puts it at $0.042 per million input tokens with output free, so 100 million input tokens is $4.20 in model charges. That figure excludes storage, orchestration and any other model you run alongside it.

What should go into the state?

The context available for the judgment, either a string or a structured JSON record. The post's advice is to include the evidence relevant to the question, not every piece of information the application happens to have.

Does a Noul come back with a confidence field?

No. Unlike Choice and Score, a Noul has no separate confidence; its value between zero and one is the signal the application uses.

How does the write-up suggest combining Jev with an LLM?

By giving each model the job it is good at: a decision model selects the route, code checks account permissions and policy, and an LLM drafts the reply.

CommunityJev Guides & Articles

TypeSafe Jev explained

Requesty's September 2026 guide to Jev as a System One model: state, Choice, Score, and Noul, plus how Requesty routes typesafe/jev-1.13.0.

See highlights

Open requesty.ai

Format: Article
Published by: Community
Author: Requesty
Last verified: Sep 2026

Tagscommunity article explainer pricing

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Highlights

Last updated 19 September 2026. Frames Jev as returning decisions, not paragraphs.
Walks Choice, Score, and Noul with illustrative probability tables, not live API results.
Lists Jev pricing at $0.042 per million input tokens with free output.
Example: 100 million input tokens at that rate is $4.20 in model charges.
Notes Jev on Requesty as typesafe/jev-1.13.0 and points at the Decisions API shape.

Watch out

Independent gateway write-up. Illustrative tables in the post are not billed API runs.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: New signups no longer get free credits.

Hacker News: Simon Willison's overview of Noul, Choice, and Score, and where a decision model fits.

benwerd53 pointsJev introduces a new shape of LLMRead the thread on Hacker News

Hacker News: swyx posts the Latent Space interview with Diogo Almeida.

swyx4 pointsJev: System One Models for Prod, Not God – With Diogo Almeida, CEO, TypeSafe AIRead the thread on Hacker News

X: Diogo Almeida in five minutes

Related terms

Glossary definitions related to this entry.

Token pricing

More Jev resources like this

Jev model explained

OmniaKey's fact-checked table for jev-1.13.0, limits, and pricing, plus a live connectivity check through OmniaKey's alpha search endpoint.