Serverless environments
Serverless environments is a Databricks compute / bi capability, introduced March 2024.
The versioned Python runtime a serverless notebook or job runs on: an environment version (1 through 6 so far) pins a Python release and a set of packages behind a stable Spark Connect client API so Databricks can upgrade the server underneath without touching your code, and a base environment layers a shareable YAML of extra dependencies on top - picked from the notebook's Environment side panel, and since September 2026 usable on classic compute too (Beta).
- Every environment version ships with a three-year expiry date printed next to it - Databricks supports a three-year lifecycle for all environment versions, so version 1 (released March 12, 2024) has an end-of-support date of March 12, 2027 already on the books.
- Version 1 ran Python 3.10.12 and version 2 Python 3.11.10, then versions 3, 4, 5, and 6 all list Python 3.12.3 - four consecutive "new" environments on the exact same interpreter.
- Since August 2026 a serverless notebook on environment version 5 or above can run %uv init, %uv add, and %uv sync to create a Python project, add dependencies, and sync its environment - a notebook that grows its own pyproject.
Limitations: Base environments: in jobs they work only for notebook, Python script, and Python wheel tasks (workspace base environments in jobs are Beta); Lakeflow pipelines do not support them; only the Python dependencies in the dependencies list are installed at runtime; environment version 1 is not supported (use 2 or higher); every base environment is visible to all workspace users; and a workspace is limited to 10 base environments. On classic compute the Environments dependency mode (Beta) does not support Spark JAR tasks or %scala and disables Docker, Spark environment variables, init scripts, and cluster libraries. Do not install PySpark, or anything that pulls it in, on a serverless notebook.
Open in REbricked →- Category
- Compute / BI
- Also known as
- environment version, base environments, serverless environment
- Verified
- 2026-09-20
Sources
- Official Official Databricks / Microsoft docs
- Official Databricks docs: Serverless compute release notes - the environment version list and every server-side version
- Official Databricks docs: Environment version release notes - release and end-of-support dates for versions 1 to 6
- Official Databricks docs: Environment version 6 - the one that turned off the Py4J gateway for dbutils and dropped /databricks/runtime/info.json
- Official Databricks docs: Configure the serverless environment - the Environment side panel, base environments, and job-task configuration
- Official Databricks docs: Manage workspace base environments
- Official Databricks docs: Manage dependencies using environments - the Environments dependency mode that brings base environments to classic compute (Beta)
- Official Databricks release notes: Serverless notebooks environment manager - the Environment side panel arrives (June 2024)
- Official Databricks release notes: Admins can now manage a workspace's serverless base environments (Public Preview, August 2025)
- Official Databricks release notes: Workspace base environments are now generally available (April 2026)
- Official Databricks release notes: Environment version 6 is now available (September 3, 2026)
- Official Databricks release notes: Base environments are now supported on classic compute (Beta, September 17, 2026)