DVC (Data Version Control) is a free, open-source Git-based tool for ML data and pipeline versioning. See features, DVC Studio pricing, pros, cons, and rivals.
Category
AI Infrastructure & MLOps
Pricing
Open Source, from Free
Verified
Not yet
Last updated
July 18, 2026
Founded
2017
Headquarters
San Francisco, California, United States (Iterative.ai); stewardship now under lakeFS
Free PlanAPIOpen SourceSelf-Hosted
Overview
DVC (Data Version Control) is a free, open-source command-line tool that extends Git's version-control model to large datasets, machine learning models, and reproducible ML pipelines.
Created by Dmitry Petrov, with the first beta released in May 2017, DVC was developed by Iterative.ai, a company Petrov co-founded with Ivan Shcheklein in 2018. In November 2025, DVC was acquired by lakeFS, which now stewards the open-source project.
Key Features
DVC tracks lightweight pointer files in Git while storing large data and model files in external storage such as S3, GCS, Azure Blob, SSH, or HDFS, letting teams version multi-gigabyte datasets without bloating their Git repository.
Its pipeline system, defined in a dvc.yaml file, expresses ML workflows as a directed acyclic graph and intelligently re-runs only affected stages, complemented by built-in experiment tracking, a lightweight model registry, and CI/CD integration through the companion CML tool.
Pricing
The core DVC command-line tool is free and Apache 2.0 licensed, with no usage limits and no account required to use it locally.
DVC Studio, the hosted visualization and collaboration layer, offers a free tier for individuals plus custom-quoted Team and Enterprise plans for larger organizations with collaboration and security requirements.
Key Features
Git-native data and model versioning — Tracks large files through lightweight pointers committed to Git, keeping data history alongside code history.
Storage-agnostic backends — Works with Amazon S3, Google Cloud Storage, Azure Blob Storage, SSH/SFTP, HDFS, or local and network paths.
Pipeline DAGs in dvc.yaml — Defines ML workflows as a directed acyclic graph and re-runs only the stages affected by a change.
Built-in experiment tracking — Tracks and compares experiment runs, with the companion DVCLive library for logging training metrics.
Lightweight model registry — Tags and promotes model versions for downstream deployment.
CML integration for CI/CD — Runs automated ML pipeline checks in GitHub Actions, GitLab CI, or other CI/CD systems.
Reproducible data lineage — Reconstructs the exact data and code combination that produced any historical result.
DVC Studio hosted layer — Provides a web interface for visualizing experiments and collaborating on ML projects.
Pros & Cons
Pros
Fully free, Apache 2.0 open-source core
Reuses the familiar Git workflow, giving it a shallow learning curve
Works with any cloud or on-premises storage backend
Strong reproducibility and pipeline dependency tracking
Cons
Large datasets still require setting up and paying for external storage