Why this exists

The problem

Big AI companies trained their models on everyone's data — for free. Now they want to lock the door behind them.

Models vanish from hosting overnight. Licenses change without warning. Entire research projects lose access to the tools they depend on.

When a model disappears, it takes the work of thousands with it.

What we do

We copy every open model we can find — weights, configs, documentation, everything.

How it's built

Three layers keep the archive alive. Kill one, the other two take over.

The catalog

The index of what exists — every model, every version, every download link. Hosted in four countries and pinned to IPFS. Take down one server, the rest answer.

The files

Every model file is hashed. The hash is its address. Copy it to a thousand servers across a hundred networks — it's still the same file. You can't seize a mathematical fingerprint.

Finding a way in

If one domain gets blocked, try another. If all four fail, use Tor. If that's blocked, use IPFS. The client keeps trying until something works.

The team

Small team. Zero revenue. No ads, no investors, no premium tier.

Open models stay open. That's the job.

Transparency

Everything is public. Infrastructure docs, signing keys, decision logs — all visible. If we're doing something wrong, you'll see it.

Public key

Every manifest we sign uses this Ed25519 key. Pin it, cross-check it, and reject any manifest that doesn't verify against it.

Ed25519 public key (hex):
7F83A2C9 4B5E6D1F 8A2C3B4E 5D6F7A8B 9C0D1E2F 3A4B5C6D 7E8F9A0B

Fingerprint: 7F83…9A0B

The same key is published on Nostr, embedded in the omm CLI, and pinned to IPFS. Verify it from multiple sources before trusting.

Help keep it alive

Every contribution stays anonymous. Every cent goes to infrastructure.