Comparing answers, or comparing ways of thinking

Model Council runs one question past several frontier models and shows where they agree, where they contradict each other, and what each adds. It is a good feature, it shipped first, and it reaches far more people than this site does. The difference is what is being compared.

The short answer

Model Council compares several models on one answer, now. This site reads a question through models that people wrote, each carrying a stated falsifier and a date — so the reading can later be shown to have been wrong.

Start with what is the same, because it is real: both put one question in front of several different readers instead of one, and both treat disagreement as information rather than noise. If you have used Model Council and found the contradictions more useful than any single answer, that instinct is the same one this site is built on.

Now the difference, which is about the readers, not the format.

Model Council's readers are frontier models — the same handful every product can call. They are excellent, general, and interchangeable: nobody authored them for a subject, and they do not persist between your questions. What you get is a snapshot of what current systems say today.

Here a reader is a document a person wrote: premises, a mechanism, at least one thing that would prove it wrong, and a deletion clause naming when its author would retire it. Someone who has watched one industry through three cycles can write one. It is theirs, it lives in their repo, and it carries a record.

That is why the two things end differently. A council session ends with a synthesis. A reading here ends with a claim that has a date on it — and on that date, the mechanism that produced the reading is shown to have held or not. Being outvoted and being wrong are different, and only the second is checkable.

One more practical difference: models here are documents, so you can read one before trusting it, disagree with its premises, fork it, or paste it into whichever assistant you already use. There is no subscription and nothing to sign in to.

Side by side

Perplexity Model Council	Model Meets Reality
Compares frontier models supplied by the vendor	Compares models written by people, owned by them
Same models available to every product	A model exists only if someone wrote it
Ends in a synthesised answer	Ends in a dated claim that later resolves
Disagreement shown between models	Disagreement shown, plus what would settle it
Polished product, wide reach	Opening small, self-graded, no audience yet

What Perplexity Model Council does better

Reach, polish, and the fact that it works today for anyone with a subscription — none of which describes this site. It also runs the strongest available frontier models, whereas a model here is only as good as the person who wrote it, and most people write a weak first one. If what you want is a fast, reliable read on a question from several strong general systems, use Model Council; that is what it is for, and this site does not replace it.

Where this stands today

Said plainly, because the whole point of this site is not overclaiming. The Model Garden is opening small — no models listed as of September 2026. Every record shown is self-graded: the author's own count of claims made and resolved, labelled as such on each card. Nothing here is ranked, and no independent resolution layer exists yet. Perplexity Model Council has things this does not, named above rather than omitted. Compare the designs, not the scoreboards — there is no scoreboard.

https://github.com/someone/their-model Help me use this

A model is a document, so it runs anywhere — including inside the assistant you already pay for.

The difference that is not on the table

Perplexity Model Council compares what several models say. This compares ways of thinking, and then checks them. Running one question past several models is genuinely useful, and they do it with a polish and a reach this site does not have. The difference is what is being compared: there, the same frontier models every product has access to, on one answer, today. Here, models people wrote — premises, a mechanism, and a stated falsifier — whose readings carry a dated claim that later turns out to have held or not.

Share your understanding →

Browse the Model Garden Publish your own