What Is the Difference Between Mamba and Micromamba?

Mamba vs micromamba explained: shared code base, static vs dynamic builds, base environments, installation and which one to choose.

Illustration comparing two boxes labelled mamba and micromamba
Illustration created for this guide

Mamba and Micromamba come from the same project, use almost the same commands and install the same conda packages. So why do two tools exist? This guide explains the real differences in plain language and helps you choose.

The short answer

Micromamba is the fully static, single-file build. It needs no base environment and ships no Python. Mamba is the dynamically linked build that is normally installed into a base environment, most commonly through the Miniforge installer. Since mamba 2.0 both are built from the same C++ code base, libmamba, and share the same command-line interface.

A little history

Mamba started as a faster drop-in replacement for conda. Early versions were a Python program that combined conda's code with a fast C++ solver. Micromamba appeared later as a tiny, separate executable for situations where installing a full Python-based tool was too heavy, such as CI jobs and Docker images.

With mamba 2.0 the project removed the Python layer. The official 2.0 change notes describe the new mamba executable as the dynamically linked version of micromamba, which means the two now have the same user interface and behaviour. The import mamba Python package was removed, while Python bindings remain available through libmambapy.

Side-by-side comparison

PointMicromambaMamba
Build typeStatically linked, single fileDynamically linked against libmamba
Base environmentEmpty; not requiredLives in a base environment
Bundled PythonNoneComes with the base environment (e.g. Miniforge)
Typical installOfficial script, Homebrew or one binaryMiniforge installer
CommandsSame CLI (create, install, activate, run...)Same CLI
Default config file shippedNone (existing .condarc/.mambarc are read)Installer usually configures conda-forge
LicenseBSD-3-ClauseBSD-3-Clause

Why the base environment matters

A base environment is a full conda environment where the package manager itself lives. It is convenient on a desktop because you get a Python interpreter and tools right away. But it also means the package manager can be affected by what you install into base, and it takes more space.

Micromamba avoids this. Its base environment is completely empty, and all environments and caches go under the folder set by MAMBA_ROOT_PREFIX. The binary itself never changes when you install packages, so it is very hard to break.

Are the commands really the same?

For everyday work, yes. These commands work the same way in both:

micromamba create -n analysis python=3.12 pandas -c conda-forge
micromamba activate analysis
micromamba install matplotlib
micromamba env list
micromamba env remove -n analysis

Replace micromamba with mamba and the result is the same. Both also offer run, which executes a command inside an environment without activating it, and shell init for activation support.

When to choose Micromamba

  • CI pipelines: download one binary, create the environment, run tests.
  • Docker images: no base Python means smaller layers.
  • Servers and shared machines: install without admin rights in your home folder.
  • Minimal laptops: you only want project environments, not a global Python.

When to choose Mamba

  • You want a ready-made base environment with Python, installed by Miniforge.
  • Your team documentation already says "install Miniforge".
  • You use tools that expect a conda-style base installation.

Can I switch later?

Yes. Because both read the same environment.yml format, you can export an environment with one tool and recreate it with the other:

mamba env export -n analysis > environment.yml
micromamba create -n analysis -f environment.yml

Keep in mind that each tool keeps its own root prefix, so environments are not automatically shared unless you point both at the same location.

Is one faster than the other?

For solving and installing packages, no meaningful difference should be expected: both run the same libmamba code. Where Micromamba feels faster is bootstrapping. Downloading one small binary takes seconds, while installing Miniforge means downloading and unpacking a full base environment first. On a CI runner that starts from a clean machine for every job, that difference adds up quickly.

Start-up time is also similar, since mamba 2.x no longer starts a Python interpreter. The old advice that "mamba is slow to start because of Python" applied to mamba 1.x and earlier.

Configuration differences to know

  • Root prefix: Micromamba uses MAMBA_ROOT_PREFIX, typically ~/micromamba. Mamba from Miniforge uses the Miniforge install folder, for example ~/miniforge3.
  • Channels: Micromamba ships no channel configuration, so you add conda-forge yourself (the install script can do it). Miniforge comes preconfigured for conda-forge.
  • Config files: both read .condarc and .mambarc, so one user configuration can serve both tools.

Common mix-ups

If you install both tools, be careful which one your shell activates. Each has its own shell init block in your profile, and the last one loaded wins. Running which mamba micromamba and micromamba info shows which binary and root prefix are in use. Many people keep only one tool on a given machine to avoid confusion.

Summary

Mamba and Micromamba are two builds of the same engine. Choose Micromamba when you want a single, self-contained binary with no base environment; choose mamba when you want a traditional conda-style installation. Ready to try it? Follow our Linux and Ubuntu install guide.

Download Micromamba 2.10.0

Official install scripts (main.zip) from GitHub, BSD-3-Clause. The download starts on this page.

Download