What is Model Registry

Definition

A model registry is a centralised repository that stores, versions, and tracks machine-learning models along with their metadata, lineage, and deployment status, providing governance and a single source of truth for which model version is where across the lifecycle.

« Back to Glossary Index

Versions every model so teams know exactly what is deployed where
Centralises metadata, metrics, and lineage for each model
Governs promotion of models through staging to production
Enables rollback to a known-good model version when needed

Real World Example

An ML team registers each trained fraud model in a model registry with its metrics and data lineage, so they can promote a validated version to production and instantly roll back to the previous one if it underperforms.

FAQs

What does a model registry store?

Model versions plus metadata such as metrics, parameters, lineage, and deployment stage, serving as the source of truth for models.

Why use a model registry?

It brings versioning, governance, and traceability to models, making deployments controlled, auditable, and reversible.

How does it fit into MLOps?

It is a core MLOps component linking training, validation, and deployment by tracking which model versions move through each stage.