Miniforge

Miniforge is a minimal installer for Conda and Mamba preconfigured for the conda-forge community package channel. It suits developers, scientific-computing users and data analysts who want isolated environments rather than modifying a system Python installation. It is a command-line environment-management distribution, not a graphical application store or a full scientific-software bundle.

Main features

• Install Conda, Mamba and their dependencies into a managed base environment.
• Create isolated project environments with their own Python and package versions.
• Resolve and install packages from conda-forge, the default and only configured channel.
• Activate environments from supported shells after initialization.
• Use native installers for multiple operating systems and CPU architectures.
• Use interactive or documented batch installation workflows, with separate macOS PKG options.

Install and start

This catalog's Mac cask can be installed with brew install --cask miniforge, but upstream explicitly does not recommend Homebrew installation: its repackaging can introduce incompatibilities that upstream does not test. If you prefer upstream's supported installer workflow, choose the official .sh or macOS .pkg matching your architecture and follow the README. Do not install competing distributions into the same prefix.

After installation, initialize the shell you use when needed (for example, conda init zsh for zsh), then restart the shell. The current cask includes a shell-initialization caveat; batch upstream installation does not initialize Conda automatically. Initialization changes shell configuration and can activate the base environment by default, changing which python runs. Create a named environment for project dependencies, such as conda create --name my_project python=3.13, then activate it with conda activate my_project. Upstream documents disabling automatic base activation with conda config --set auto_activate_base false.

Manage packages through the environment tools deliberately; mixing unrelated distributions, global PATH changes and overlapping prefixes can cause conflicts. Before uninstalling, review environment data and shell initialization rather than deleting an uncertain directory.

Versions, packages and channels

The current 26.7.2-0 installer recipe pins Conda 26.7.2, Mamba 2.9.0, conda-libmamba-solver 26.7.0 and Python 3.14 in base. This installer release identifier is not the version of every bundled component. Version 26.5.3-0 changed base Python from 3.13 to 3.14; 26.3.2-3 was the last release with Python 3.13 in base. Other environments can use different supported Python versions and implementations, so base Python does not lock every project to 3.14.

Base packages come from conda-forge and the configured channel is conda-forge, not Anaconda's default channel. The recipe also defines conda-forge mirrors; this does not imply that every custom channel has the same terms or package availability. Additional tools such as JupyterLab, NumPy and pandas are installed into chosen environments as needed rather than assumed to be bundled.

Requirements, dependencies and platforms

Current official installers require macOS 11 Big Sur or newer for Intel x86-64 and Apple Silicon ARM64. The README labels Apple Silicon and Linux RISC-V builds experimental with less testing. Linux x86-64, ARM64 and POWER 8/9 ppc64le require glibc 2.17+; the new Linux riscv64 target requires glibc 2.29+. Windows x86-64 requires Windows 10+; upstream does not establish the exact earliest Windows 10 build. A 64-bit Raspberry Pi installation needs a 64-bit OS. No 32-bit/mobile installer is established here.

The Homebrew cask conflicts with miniconda; select one rather than installing both casks together. A separate system Python, Homebrew Python formula or compiler is not a universal requirement for the bundled installer. Some packages need their own hardware, operating-system features or external dependencies. No universal RAM or installed-space minimum was confirmed: installed environments and package caches determine storage needs.

The official 26.7.2-0 Apple Silicon shell installer is 83,747,533 bytes from GitHub release asset metadata, matching the production package; this is transfer size, not installed space. The installer includes bootstrap/base content, while fetching new packages needs network access unless you intentionally provide an offline package setup. No account is required for ordinary public conda-forge environment use; private channels and external services may require their own credentials.

Price and license

Miniforge is free and its installer code uses BSD 3-Clause. Bundled packages and the temporary bootstrap executable have their own licenses; review the installer license and individual package notices, especially for redistribution. Downloading Miniforge does not grant a blanket license to all possible third-party packages or paid services.

Sources: Official README and installation caveats, Current release, Version-specific recipe, License, conda-forge installer overview.