#Share Models

#Overview

Model sharing is handled through Model Catalog sources and Model Registry permissions.

Use Model Catalog when an administrator wants to make models from an upstream source discoverable to users. Use Model Registry when a team needs to share registered models, versions, metadata, and artifact locations with other users or namespaces.

A shared model still needs a reachable artifact location. The Model Catalog or Model Registry stores the model metadata and Model URI; the model bytes remain in S3, a PVC, or an OCI registry. Configure the corresponding Connection, Secret, or service account so the inference service can access the artifact.

#Share models through Model Catalog

#Configure a catalog source

An administrator can add a source from Admin View > Settings > Model Catalog Sources:

1. Select Add Source.
2. Enter a source name using lowercase letters, numbers, and underscores.
3. Select the source type:
   • Hugging Face: enter the required Organization and, if required, an Access Token.
   • YAML: enter the catalog YAML content.
4. Optionally set Display Name, Allowed Models, Excluded Models, and Enabled.
5. Select Add and wait for the source status to become available.

Allowed and Excluded Models can limit which models are exposed by that source. Do not include credentials in YAML content or model metadata. Edit or delete a source from its row action menu when the source is no longer needed.

#Browse and use a catalog model

A user can open User View > AI Hub > Model Catalog, choose a source, and open a model entry. The details page can show the provider, description, model card, task type, labels, framework, license, size, and artifact URI.

From the details page, the user can:

• select Create Inference API to deploy the catalog artifact directly; or
• select Register Model, choose an available Model Registry, and create a local registered model and version that references the catalog artifact.

Catalog entries themselves are read-only. To edit the local metadata or control its lifecycle, register the model and manage the resulting Model Registry record.

#Share models through Model Registry

#Grant access to a Model Registry

An administrator can manage registry access from Admin View > Settings > Model Registries:

1. Find the target registry.
2. Open its row action menu and select Manage permissions.
3. Add the required users, groups, or namespaces and assign the appropriate permission.
4. Save the permission changes.

Adding a namespace allows service accounts in that namespace to access the registry according to the configured permission. Grant only the access needed for the team or workload.

#Manage a registered model

A user with the required registry permission can open User View > AI Hub > Model Registry and:

1. Select the Model Registry containing the model.
2. Open the registered model to review its Overview and Versions tabs.
3. Update labels, descriptions, or custom properties when needed.
4. Open a version to inspect its Model URI, framework, state, and related inference services.
5. Select Create Inference API or Deploy for a LIVE version.

The model and version metadata are maintained by Model Registry. They are independent of the catalog source metadata.

#Archive and restore

Archiving removes a model or version from the active view while retaining its metadata and artifact reference:

1. Open the model or version action menu and select Archive.
2. Type the exact model or version name in the confirmation field.
3. Select Archive.
4. Use the archived view and Restore action to make it active again.

A model or version that is used by an inference service cannot be archived. Delete the related inference service first. Archiving a registered model also archives its versions. An archived entry is not available for a new deployment until it is restored.

#Recommended sharing patterns