In plain terms

Get AI models from infrastructure Europe runs itself. With proof that what you received is genuine.

An AI model is a file someone has to hand you. acervus is an open model registry being built so that any node can serve the same files, and any machine can check for itself what it received: the exact bytes, where they came from, and under which licence. It is designed to work with the tools people already use: point one environment variable at a node and the existing clients work unchanged.

Registry Mirror nodes You Verify offline, no connection needed
Where this stands · 11 September 2026

This page will change as things become true. It will not say they are true before that.

What it is
  • Artefacts served byte-identical to their upstream source, named by their own fingerprint.
  • Origin, licence and version history signed and travelling with the files.
  • Verification anyone can run, without needing the registry to be online.
  • Compatible with existing AI clients; nothing to re-learn.
What it is not
  • Not a place to discover, rank or discuss models.
  • Not hosted inference, training or a marketplace.
  • Not a replacement for the platforms people already use; it is the layer underneath, and it can mirror them.
Contact

Questions, institutions interested in running a node, or publishers who would like to sign their releases: hello@acervus.eu

Built in Prague. Independent. No funding, partners or customers to name yet.