AI Potluck
Contribute
Back to Gap Map Infrastructure / Model hubs & registries

Kubeflow Hub (Model Registry)

Kubeflow
open source / Overall score: n/a

Kubeflow component, formerly called Model Registry, where teams register models, model versions and their artifacts through a REST API and Python client and track them from experimentation to production. The registry stores metadata and the URI of each artifact in S3 or an OCI registry, and its client can upload a local model and register it in one call. A companion catalog lets users browse models from external sources such as Hugging Face before registering them. The component is still marked alpha.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(github.com/kubeflow/hub)
core features withheld
no — a component of Kubeflow, a CNCF graduated project

The registry and catalog are Apache-2.0 and install standalone or with Kubeflow. Red Hat drives its development, but the repository carries no enterprise code or license check.

Adoption

not assessed

Kubeflow Hub runs inside each organization's own cluster, and no install or user count is published. Downloads of its Python client would measure a client rather than the registry, so no reading was recorded.

Capability

4 medium confidence
4.0

Team members register models, versions and artifacts through a documented API, and the client uploads the files to the deployment's own object store or OCI registry as it registers them. It tracks models only and keeps weights in the storage it points to rather than serving them itself, which keeps it a step below Hugging Face Hub.

Verified 2026-09-27