Pagouro

A language model on a USB stick that says when it does not know.

A one-billion-parameter model, trained from random weights on a corpus where every byte has a licence you can name and a date before 2022, packaged with everything it needs. It runs on any CPU, offline, with no account. It is about a thousandth the size of the models you already use and loses to them on every capability test. What it has instead is four promises a stranger can check — and the checks ship with it.

13 %of unanswerable questions get an invented answer (the bluff rate, on a frozen 100-item test; small open models 50–57 %, frontier models 23–27 %)

82 %of real questions answered correctly (the matching 100-item set). Both numbers always together: a model that says nothing would ace the first alone.

The four promises

Licensed. Every training source is a row in corpus.json — licence, date basis, token count, hash — including the rows that were removed and why.
Dated. Everything it read was written or collected before 1 January 2022, per row.
Honesty measured. The two numbers above, on a test written before the model existed. The test is in the repository; run it on anything.
Finished. One release, signed, its hash timestamped on Bitcoin, mirrored. No updates, no telemetry, no watermark, no claim on what it writes for you.

What it is not

Not a frontier model, and not trying to be one. It reasons badly (11 of 320 school word problems without tools — that number is on the box too), writes short, and points arithmetic at a calculator. If you want a frontier model, use one. If you want to know exactly what you are talking to, this is the niche, and this is the first thing in it.

Get it, check it

Download — the release zip, with its SHA-256 beside it (at release) Weights and the corpus ledger on Hugging Face (at release) Permanent copy on Arweave (at release) Source, evals, the decision log, the book — the repository (public at release)

Manifest SHA-256: 02a618fc07bdfdc84b40eefe28cbf6080dfba856dee7eef2c8164223abd90f71
Signing key (minisign): RWQe8tvI6RCE2uMbuILC9/rEr6bNZdcOA+WC7dHtObLE94ovGk8xuFlG
Bitcoin timestamp (OpenTimestamps): at release
In the folder, python verify_manifest.py must print every listed file matches the manifest.