ModelDB is an open-source system for machine learning model versioning, metadata tracking, and experiment management, created by Verta. It has 1,752 stars on GitHub and was last updated in 2024. ModelDB treats models like code under version control, recording everything needed to reproduce them.
When a model performs well, teams need to know exactly which code, data, configuration, and environment produced it. ModelDB captures those ingredients automatically across the model lifecycle, from experimentation through deployment and live monitoring. It offers Git-like operations on models, flexible logging of metrics, artifacts, and tags, and dashboards for comparing performance across runs. The system runs on Docker or Kubernetes, offers Python and Scala clients, and integrates with frameworks such as TensorFlow and PyTorch with pluggable storage backends. ModelDB began as a research system from MIT and was later developed by Verta as an open-source project. Every logged version is immutable, so an experiment can be re-opened months later with its exact ingredients intact, which is essential for audits, hand-offs between team members, and regulated industries.
ModelDB is free and open source under the Apache-2.0 license for self-hosting. Verta historically offered hosted and enterprise options for teams that wanted a managed deployment.
ModelDB is for ML teams that need reproducibility and auditability: data science groups versioning many experiments, and platform teams standardizing how models are tracked into production. Its longevity also makes it a useful reference point for how the model registry category has evolved.