Skip to content

mamba

mamba is a fast, drop-in replacement for conda. It uses the same commands but resolves dependencies much faster.

Mental Model

mamba is conda rewritten in C++ for speed. Every conda command you know works identically with mamba — just swap the word. The dramatic speedup comes from a faster dependency solver (libmamba), which matters most when you have large environments with many packages competing for compatible versions.

Why mamba?

conda's dependency solver can be very slow, especially with:

  • Many packages
  • Complex dependency trees
  • Large environments

mamba solves the same problems 10-100x faster using:

  • C++ implementation (libsolv)
  • Parallel downloads
  • Better dependency resolution algorithm

Installation

Fresh installation with mamba pre-installed:

```bash

Linux

wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-Linux-x86_64.sh bash Mambaforge-Linux-x86_64.sh

macOS (Apple Silicon)

wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-MacOSX-arm64.sh bash Mambaforge-MacOSX-arm64.sh

macOS (Intel)

wget https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-MacOSX-x86_64.sh bash Mambaforge-MacOSX-x86_64.sh ```

Option 2: Install mamba in Existing conda

bash conda install -c conda-forge mamba

Usage

mamba uses identical syntax to conda. Just replace conda with mamba:

Package Management

```bash

Install packages

mamba install numpy pandas scikit-learn

Install specific version

mamba install numpy=1.24

Update package

mamba update numpy

Remove package

mamba remove numpy

Search packages

mamba search numpy ```

Environment Management

```bash

Create environment

mamba create -n myenv python=3.11 numpy pandas

Activate (still use conda)

conda activate myenv

Install in environment

mamba install matplotlib

Deactivate (still use conda)

conda deactivate ```

Note: Use conda activate/deactivate — mamba doesn't replace these commands.

Environment Files

```bash

Create from file

mamba env create -f environment.yml

Update from file

mamba env update -f environment.yml

Export (use conda)

conda env export > environment.yml ```

Speed Comparison

Operation conda mamba
install numpy pandas scikit-learn 30-60s 3-5s
Create environment with 50 packages 5-10 min 30-60s
Resolve complex dependencies Minutes Seconds

The difference is most noticeable with:

  • Many packages
  • Version conflicts to resolve
  • Fresh environments

When to Use mamba vs conda

Task Use
Install packages mamba install
Create environments mamba create
Update packages mamba update
Search packages mamba search
Activate environment conda activate
Deactivate environment conda deactivate
Export environment conda env export
Config changes conda config

Common Commands

conda mamba equivalent
conda install pkg mamba install pkg
conda create -n env mamba create -n env
conda update pkg mamba update pkg
conda remove pkg mamba remove pkg
conda search pkg mamba search pkg
conda env create -f file mamba env create -f file
conda clean --all mamba clean --all

Micromamba

micromamba is an even smaller, standalone version:

  • No base environment needed
  • Single static binary (~5MB)
  • No Python dependency
  • Great for CI/CD and containers

Install micromamba

```bash

Linux/macOS

"${SHELL}" <(curl -L micro.mamba.pm/install.sh)

Or download directly

wget https://micro.mamba.pm/api/micromamba/linux-64/latest -O micromamba chmod +x micromamba ```

Use micromamba

bash micromamba create -n myenv python=3.11 numpy micromamba activate myenv micromamba install pandas

Troubleshooting

mamba Not Found After Installation

```bash

Restart shell or source config

source ~/.bashrc # Linux source ~/.zshrc # macOS zsh

Or specify full path

~/mambaforge/bin/mamba install numpy ```

Still Slow?

Make sure you're using mamba, not conda:

bash which mamba mamba --version

Conflicts with conda

If you have both conda and mamba, they share the same environments. No conflict, but be consistent.

Best Practices

1. Use Mambaforge for New Setups

bash bash Mambaforge-Linux-x86_64.sh

Gets you mamba + conda-forge by default.

2. Create Alias (Optional)

If you always want to use mamba:

```bash

In ~/.bashrc or ~/.zshrc

alias conda='mamba' ```

Then conda install actually runs mamba install.

3. Use mamba for Heavy Operations

  • Installing multiple packages
  • Creating environments
  • Updating environments

Use conda for:

  • activate/deactivate
  • config changes
  • env export

Summary

Tool Description Speed
conda Original package manager Slow
mamba Fast conda replacement Fast
micromamba Minimal standalone mamba Fast

Key Takeaways

  • mamba is a drop-in replacement for conda
  • Same syntax: mamba install, mamba create, etc.
  • 10-100x faster than conda for dependency resolution
  • Use Mambaforge for new installations
  • Still use conda activate/deactivate
  • micromamba for containers and CI/CD

Exercises

Exercise 1. Write the mamba commands to: (a) create a new environment called ml-env with Python 3.11, numpy, and scikit-learn, (b) activate it, and (c) install an additional package pandas. Explain which commands still use conda and which use mamba.

Solution to Exercise 1

```bash

(a) Create environment (use mamba)

mamba create -n ml-env python=3.11 numpy scikit-learn

(b) Activate (still uses conda)

conda activate ml-env

(c) Install additional package (use mamba)

mamba install pandas ```

Use mamba for install/create/update (faster dependency resolution). Use conda for activate/deactivate (environment management).

Exercise 2. Explain the key difference between mamba and micromamba. When would you choose micromamba over mamba, and what are its limitations?

Solution to Exercise 2

mamba is a drop-in replacement for conda written in C++. It requires an existing conda installation (or Mambaforge) and shares conda's environment management.

micromamba is a standalone, statically linked binary that does not require conda or Python. It is ideal for:

  • Docker containers (small image size)
  • CI/CD pipelines (fast setup, no conda dependency)
  • Minimal installations

Limitations of micromamba:

  • Does not support all conda commands
  • No conda activate -- uses micromamba activate or shell eval
  • Smaller community and fewer tutorials

Exercise 3. Write a Dockerfile snippet that uses micromamba to create a Python environment and install packages. The environment should have Python 3.11 and Flask.

Solution to Exercise 3

```dockerfile FROM mambaorg/micromamba:latest

Create environment

RUN micromamba create -n app python=3.11 flask -c conda-forge -y

Activate environment in subsequent commands

ARG MAMBA_DOCKERFILE_ACTIVATE=1 ENV ENV_NAME=app

WORKDIR /app COPY . .

CMD ["python", "app.py"] ```