Skip to content
BSWEN
Blog

Conda vs UV vs Poetry: When Should You Choose Which Package Manager?

Cowrie
Cowrie
Dev @ Bswen

Conda vs UV vs Poetry: When Should You Choose Which Package Manager?

I’ve been working with Python for years and I’ve faced this question countless times: which package manager should I use? The debate between Conda, UV, and Poetry comes up constantly. I tried all three and I got frustrated with each one in different scenarios. The key insight I discovered is that no single tool is best for everything - each has strengths for different project types.

The Problem

I tried using Conda for a simple web project once and it was a nightmare. Slow dependency resolution, confusing channels, and unnecessary complexity. But then I tried using UV for a machine learning project and struggled with missing C++ dependencies. I think the confusion comes from not understanding what each tool is designed to solve.

Choose the Right Tool for Your Project

Here’s what I learned after testing all three across different scenarios:

Conda - Scientific Computing Powerhouse

When to use it: Scientific computing, machine learning, data analysis, any project with C/C++ dependencies

Why it excels: Conda provides pre-compiled binaries for complex dependencies that would take hours to compile from source.

Terminal window
# Create environment with specific Python version
conda create -n ml-project python=3.11
conda activate ml-project


# Install PyTorch with CUDA support - this just works
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia


# Install scientific packages optimized for performance
conda install pandas numpy scikit-learn jupyter matplotlib

The trade-off: Conda’s dependency resolution is notoriously slow. I waited 20 minutes for environment creation once. But for complex scientific workflows, it’s worth it.

UV - Speed Demon for Pure Python

When to use it: Pure Python web apps, CLI tools, projects prioritizing speed and pip compatibility

Why it excels: UV is incredibly fast. It’s pip-compatible but resolves and installs packages 10-100x faster.

Terminal window
# Install UV once
pip install uv


# Initialize project (creates pyproject.toml with dev dependencies)
uv init


# Add dependencies - this is blazing fast
uv add requests fastapi uvicorn gunicorn python-dotenv


# Run your application
uv run python main.py


# Create virtual environment
uv venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows

The catch: UV doesn’t handle binary dependencies well. If you need packages with C/C++ extensions, you’ll need Conda or pip with compilation tools.

Poetry - Elegant Project Management

When to use it: Django/Flask APIs, library development, projects requiring publishing to PyPI

Why it excels: Poetry provides beautiful dependency management and automated publishing workflows.

Terminal window
# Install Poetry
pip install poetry


# Create new project
poetry new web-api
cd web-api


# Add dependencies with automatic lockfile generation
poetry add django djangorestframework python-decouple whitenoise


# Development dependencies
poetry add --group dev pytest black isort mypy


# Run development server
poetry run python manage.py runserver


# Publish to PyPI when ready
poetry publish

The benefit: Poetry’s pyproject.toml format is clean and standardized. The automatic lockfile generation solves dependency conflicts permanently.

Comparison Table

Project TypeBest ChoiceWhySpeedDependency Complexity
Scientific/MLCondaPre-compiled C/C++ binariesSlowVery High
Pure Python Web AppUVPip-compatible speedVery FastLow
Django/Flask APIPoetryProject management + publishingMediumMedium
Data AnalysisCondaPandas/NumPy optimizationSlowHigh
CLI ToolsUVFast development iterationFastLow
Library DevelopmentPoetryPublishing to PyPIMediumMedium

Common Mistakes I Made

  1. Using Conda for simple projects: I created a basic Flask app with Conda once and regretted the unnecessary overhead.

  2. Choosing UV for complex dependencies: I tried to use UV for a project with OpenCV and NumPy - it failed because of binary dependencies.

  3. Overlooking Poetry’s publishing: I manually uploaded packages to PyPI for months before discovering Poetry’s one-command publishing.

The Real Question

The debate isn’t which tool is “best” - it’s which tool is right for your specific use case. I think the Reddit discussion got it right: Conda has complex dependency resolution but provides pre-compiled binaries, while UV and Poetry offer faster solutions for pure Python projects.

When I’m working on machine learning models, I reach for Conda. For web APIs, I use Poetry. For quick scripts and prototypes, UV is my go-to. Each tool solves different problems well.

Conclusion

Choose Conda for scientific computing with complex dependencies, UV for fast pure Python development, and Poetry for elegant project management and publishing. The key is understanding what each tool excels at and matching it to your project’s needs.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!

Comments

Tags: