Question 1

What is this, code or a list?

Accepted Answer

A list, plus the public surface for editing it. The repository is described as the contribution surface for modelsystem.one — issues, pull requests and discussions — while the website application source lives in a separate private repository. So there is nothing to install, run or self-host, and the README warns against opening pull requests that expect the private application to be here. What you get is CONTRIBUTING.md, an issue template, and the published spec the catalog applies.

Question 2

What actually qualifies a model for the catalog?

Accepted Answer

Three criteria, summarized in the README and detailed on the spec page: the primary interface must return typed decisions rather than free text, the model must be publicly available at GA, beta or waitlist, and there must be at least some API documentation. Corrections to specs, pricing, status or links are also accepted contributions, as are documentation, learn and guide improvements and bug reports against the public site.

Question 3

Which models are listed right now?

Accepted Answer

The README names one: Jev from TypeSafe AI, described as leading the public category, with the catalog's stated purpose being to track every entrant as the category grows. Anything beyond that is browsable on the site's own catalog and quadrant pages rather than in the repository, since the repo carries no model list. Treat the single-name claim as the state of the README, not an exhaustive count.

Question 4

Why should I care about typed decisions over chat?

Accepted Answer

That is the site's organizing argument, framed with the fast-and-slow distinction: traditional LLMs return free text in seconds and suit generation and reasoning, while decision models return a typed decision with a probability in milliseconds and suit classification, routing, scoring and guardrails. Worth being clear that this is the catalog's framing, not a measurement it publishes — there is no benchmark here to check it against, which is also why the inclusion criteria stop at interface shape, availability and documentation.

Question 5

What does the quadrant and learn section add that the catalog does not?

Accepted Answer

The quadrant page places models on maturity against capability, and the learn section applies the System 1 versus System 2 distinction to AI rather than leaving it as a slogan. The framing comes from Kahneman's Thinking, Fast and Slow: traditional LLMs return free text in seconds and suit generation and reasoning, while decision models return a typed decision with a probability in milliseconds and suit classification, routing, scoring and guardrails.

Question 6

What license covers this, and what happens to the vendors' names?

Accepted Answer

MIT over the repository's own content. Site content and third-party trademarks stay with their owners, and the catalog states plainly that ModelSystem.One is independent and not affiliated with the vendors it lists. Topics are suggested for discoverability — ai, system-one, decision-models, machine-learning, llm, catalog, typed-decisions, calibrated-probability, developer-tools, modelsystem — and there is a full Portuguese mirror at modelsystem.one/pt-br.

Question 7

How do I submit something, and what should I read first?

Accepted Answer

Read CONTRIBUTING.md, then open an issue from the template chooser, or send a pull request against this repository for content and meta changes only. Accepted contributions are new models meeting the catalog spec, corrections to specs, pricing, status or links, documentation and guide improvements, and bug reports about the public site. The README is explicit that pull requests expecting the private application source will land nowhere, because that source lives elsewhere.

Modelsystem

About this project

Highlights

Questions about this entry

More Jev resources like this

From the Jev guides

From the Jev community