uv vs venv vs conda: Python Environments Compared

uv, venv and conda all create isolated Python environments, but they do different jobs. See what each does, the commands, and which to choose.

For ordinary Python projects, use uv. If you want zero extra tools, use the built-in venv. Choose conda when you need non-Python software installed alongside Python, which is common in scientific and data work. All three give you an isolated environment so one project’s packages don’t clash with another’s. They differ in how much else they do.

At a glance

venv: built in and manual

venv is part of Python itself. It creates a lightweight environment on top of an existing Python installation, with its own independent set of packages.

python -m venv .venv
source .venv/bin/activate
python -m pip install requests

You activate it, install with pip, and keep your own record of dependencies. It does nothing about which Python version you have: it uses the one you ran it with. Choose it when you want no extra software and the project is simple.

uv: one fast tool for the whole workflow

uv creates environments, and it also handles Python versions and dependencies:

uv venv                       # create .venv
uv venv --python 3.11         # request a version
uv run python app.py          # run, with the project locked and synced first

If the requested Python version isn’t on your machine, uv downloads it. In a uv project, locking and syncing are automatic: when you use uv run, the project is locked and synced before your command runs, so the environment stays in step with the lockfile. That single tool replaces a combination of pip, a version manager and virtualenv for most workflows, which is the main reason it has taken off. See uv and, for the packaging choices around it, Poetry vs uv vs Hatch.

Choose uv for web apps, command-line tools, libraries and most day-to-day projects. Its limit is that it installs Python packages, so it doesn’t provide native libraries that aren’t bundled inside Python wheels.

conda: environments plus non-Python software

conda is a separate package manager. It treats Python as one package among many and can install the compiled libraries and tools around it, including software that isn’t Python at all.

conda create --name myenv python=3.12
conda activate myenv

That is why it dominates data science and scientific computing, where packages depend on specific C, C++ or Fortran libraries. It is a larger tool than uv, with its own channels and installers to understand, and many people prefer a community-run installer over the full Anaconda distribution. Our virtual environments guide covers alternatives such as Miniforge, micromamba and pixi.

How to choose

• Learning Python, or a small script: venv.
• Most new projects, web apps, CLIs, libraries: uv.
• Scientific stack with heavy native dependencies, or a course that uses conda: conda.
• A team already on one of them and happy: stay put. The best environment tool is the one everyone on the project uses the same way.

Common mistakes

1. Installing into the system Python. Always create and activate an environment first, or use uv run.
2. Committing the environment folder. Add .venv/ to .gitignore and commit your dependency list or lockfile instead.
3. Mixing managers in one environment. Installing with pip or uv into a conda environment can leave it in a state neither tool understands. Pick one tool per environment.