Pixi - Package Management That Doesn't Suck

If you've ever had conda fuck up your base environment for the fifth time, or waited 20 minutes for conda to "solve environment" only to fail, you'll love Pixi. It's made by the prefix.dev team (the same people who created mamba) and it's basically conda that actually works.

Built in Rust using the rattler library, Pixi uses the same conda ecosystem you're used to but without the soul-crushing slowness. It pulls from conda-forge and PyPI so you get access to the same 20,000+ packages, just installed in seconds instead of decades.

How It Actually Works

Pixi is a single binary - no Miniconda installation, no base environment to corrupt, no "conda activate" bullshit. Remember when you accidentally installed something in your base environment and broke everything? We've all been there - nothing quite like that sinking feeling when python stops working because you fucked up your conda base. That can't happen with Pixi.

Everything lives in a .pixi/ folder in your project. Each project gets its own isolated environment. No more "works on my machine" because Chad installed a different numpy version globally.

It generates actual lockfiles (like a real package manager should). Latest stable version as of 2024 is around v0.28.x, which uses the uv resolver for PyPI packages. This means dependency resolution that actually finishes instead of hanging forever.

Performance: It's Actually Fast

Environment creation that used to take 5+ minutes with conda now takes 30 seconds. Official benchmarks show Pixi is 10x faster than conda and 3x faster than micromamba.

I timed it myself: creating a data science environment with pandas, scikit-learn, and jupyter - conda took forever (I gave up at 6 minutes), Pixi was done before I could finish my coffee. Your CI/CD builds will actually finish before the heat death of the universe.

Tasks and Global Installs That Don't Break

You can define tasks in pixi.toml like npm scripts. Running pixi run test executes your test command in the right environment automatically - no activation step needed.

Global installs actually work properly. Want to install ripgrep or bat globally? pixi global install ripgrep puts it in its own isolated environment. No more global pollution, no more dependency conflicts when you install CLI tools.

If you're still skeptical, the FAQ below addresses the questions everyone asks when they first hear about Pixi.

Pixi is made by the same team behind mamba and built on the rattler library, so they know conda's pain points intimately. It's gained solid traction on GitHub for a tool that's still relatively new - definitely not some random weekend project.

What Makes It Fast

The core is rattler-solve, a Rust rewrite of conda's solver that doesn't suck. For PyPI packages, it uses uv from Astral (the same people making Python tooling blazing fast).

Rattler Architecture: The rattler ecosystem consists of multiple Rust crates that handle different aspects of conda package management - from parsing package metadata and solving dependencies to downloading and extracting packages. This modular design allows Pixi to parallelize operations that conda does sequentially, leading to massive speed improvements.

The big performance win came from sharded repodata in 2024, which can speed up complex environment solving by 50x. This is why your 30-package data science environment installs in seconds instead of minutes. Sometimes it's blazing fast, sometimes it hangs on one package for mysterious reasons - but even the slow times are faster than conda's best days.

CI/CD and Real-World Usage

There's an official GitHub Action that caches environments properly, so your CI doesn't waste 10 minutes reinstalling the same packages every run. The pixi-build ecosystem lets you build packages if you're into that, including CMake integration.

Community Reality

The Discord server is actually active, and the maintainers respond quickly. They do bi-weekly releases so bugs get fixed fast. Some cool stuff happening: conda-lock might officially endorse it, and people are asking for PyCharm integration. Fair warning: the Discord is helpful but everyone assumes you know Rust, so sometimes you get answers like "just check the rattler-solve crate" when you just wanted to install pandas.

There's experimental WebAssembly support for running scientific code in browsers, which is pretty neat. Also supply chain security through Sigstore if you're paranoid about package integrity.

What's Coming Next

The roadmap includes better build system integration, more language support, and team collaboration features. Since it's built on rattler, it benefits from all the conda ecosystem improvements happening behind the scenes.

The Bottom Line: Package Management That Actually Works

Remember that promise from the beginning - conda replacement that's 10x faster and won't corrupt your base environment? Pixi delivers on both counts. The team behind mamba has taken everything they learned about conda's pain points and built something that just works.

If you're doing scientific computing, data science, or any multi-language development, Pixi eliminates the friction that's been slowing you down. No more coffee breaks during environment creation. No more "works on my machine" debugging sessions. No more corrupted base environments that ruin your Monday.

Ready to get started? The links below will get you from zero to productive in under 10 minutes.