Decision Models

Built with decision models

More than 700 public repositories mentioning Jev appeared in the three days after launch. Ten of them, each checked against its own repo.

Building something? Send it to if@decisionmodels.ai and it goes on the list. Inclusion is not endorsement, and nobody pays to be here.

What is a decision model?

Decision model is the term this site uses for a system whose primary output is a bounded, machine-consumed judgement: a value from a declared answer space, usually with a score over the options. The implementation varies. It might be a classifier head, a direct readout of a language model’s logits, constrained decoding into a schema, or a purpose-trained model such as Jev.

The interface is not new. Function calling and strict structured-output APIs have been able to return schema-valid objects since 2023, and classifier heads are far older than that. What differs between the two paths below is how the answer is produced, not whether software can get a typed one. The open question is whether decision-specific training and serving make that materially better on latency, reliability or calibration.

A worked example

A customer message arrives, and the system needs a category — nothing more:

message: "My card was charged twice."

schema:
  category: ["billing", "technical", "sales", "other"]

A decision model returns the answer in the shape the software asked for. Illustrative output, not a measured model result:

{"category": {"value": "billing", "probability": 0.96}}

The value is already bounded by the declared answer space. Whether the probability beside it is useful is a separate question: a normalised score is not a calibrated one, and calibration has to be measured against representative data.

This did not start in September

When Jev launched, one of the loudest replies came from a GLiNER2 co-author, pointing out that scoring declared options in a single forward pass has had open weights for years. He is right, and it is worth being precise about what is actually new.

1. Loading…

What Jev claims that this lineage does not is calibration as an explicit training objective, plus the latency and price bundle. Those are vendor-reported and not yet independently reproduced, which is the gap this site intends to close.

Ecosystem

Jev entered early access on 15 September 2026, and the community projects below appeared in the days after it. Entries are marked as hosted models, local experiments, model adapters or tooling, because they are not peers. Claims are sourced to maintainer documentation and marked vendor-reported where nobody has independently reproduced them. Unknown fields are left out rather than guessed. Inclusion is not endorsement, and this is not a complete inventory of classifiers, structured-output models or decision systems.

One hosted model, four things you can run yourself, one router. The detail for each is below.

What should we call these models?

The category name is not settled.

Where each one comes from, and what it gets right and wrong:

System One Models

TypeSafe AI's own term for Jev, borrowing the fast-versus-deliberative distinction popularised by Daniel Kahneman. The System 1 and System 2 labels predate his book; Keith Stanovich and Richard West were using them in 2000. Precise, but vendor-specific — and not the only phrase TypeSafe uses. Their own AI primer describes the work as training “decision models with calibrated probabilities instead of optimizing for generated text.”

Decision Models

A broader descriptive term for systems whose primary output is a bounded judgement rather than text. Vendor-neutral, but not semantically empty: "decision model" already means something specific in decision analysis, operations research, and the DMN standard.

Probabilistic Decision Models

More technically precise: a distribution over a bounded set of choices. Narrower, and it invites a claim about calibration that usually has not been measured.

This site uses decision models as a descriptive umbrella for learned systems whose primary output is a bounded judgement. That is a practical choice, not a verdict: TypeSafe calls Jev a System One Model.

Footnote. “Reflex model” is an older adjacent term in AI teaching for a predictor that maps input straight to output with no deliberation. It describes the same shape, but there is no evidence it is in use as a name for the current Jev-related projects, so it is not listed above.

The two paths, axis by axis

Generalisations, not laws: capabilities vary by model, and the two approaches are complementary rather than mutually exclusive.

Use cases

Show all ten

Resources

Follow the decision-model ecosystem.

New models, benchmarks, APIs and open-source projects.

We store your email address and the time you signed up, in a Cloudflare D1 database. Nothing else. No advertising, and you can ask to be removed at any time by replying to any email or writing to if@decisionmodels.ai.