Conda & Anaconda Cache Cleanup on Mac — Reclaim 10-40GB+

Updated March 2025 -- Covers Anaconda, Miniconda, Miniforge, Mamba, and conda-forge -- 10 min read

If you use conda for data science, machine learning, or scientific computing on macOS, you may not realize how much disk space your conda installation is consuming. Between downloaded package tarballs, extracted packages in pkgs/, multiple environments, and the base installation itself, it's common to find 10-40GB+ of conda-related files on a developer Mac.

This guide covers every conda-related disk consumer on macOS, with exact commands to audit and clean each one safely.

1. Quick Audit — Where Is Your Conda Disk Space Going?

Before cleaning anything, let's see exactly what's consuming disk space. Run this audit script:

# Full conda disk audit
echo "=== Conda base installation ==="
du -sh ~/miniconda3/ 2>/dev/null
du -sh ~/anaconda3/ 2>/dev/null
du -sh ~/miniforge3/ 2>/dev/null
du -sh ~/mambaforge/ 2>/dev/null
du -sh /opt/homebrew/Caskroom/miniconda/ 2>/dev/null

echo ""
echo "=== Package cache (pkgs/) ==="
du -sh ~/miniconda3/pkgs/ 2>/dev/null
du -sh ~/anaconda3/pkgs/ 2>/dev/null
du -sh ~/miniforge3/pkgs/ 2>/dev/null

echo ""
echo "=== Environments ==="
du -sh ~/miniconda3/envs/*/ 2>/dev/null | sort -hr
du -sh ~/anaconda3/envs/*/ 2>/dev/null | sort -hr

echo ""
echo "=== Conda metadata ==="
du -sh ~/.conda/ 2>/dev/null
du -sh ~/.condarc 2>/dev/null

echo ""
echo "=== Total conda disk usage ==="
du -sh ~/miniconda3/ ~/anaconda3/ ~/miniforge3/ ~/mambaforge/ 2>/dev/null

Typical results on a data science Mac:

2. Package Cache (pkgs/) — The #1 Space Consumer

The pkgs/ directory is conda's package cache. Every package you've ever installed is cached here — both the compressed tarball and the extracted version. This directory only grows; conda never automatically cleans it.

# See exactly what's in the package cache
du -sh ~/miniconda3/pkgs/ 2>/dev/null
ls ~/miniconda3/pkgs/ | wc -l

# Top 20 largest cached packages
du -sh ~/miniconda3/pkgs/*/ 2>/dev/null | sort -hr | head -20

# Count compressed tarballs vs extracted dirs
ls ~/miniconda3/pkgs/*.tar.bz2 2>/dev/null | wc -l
ls ~/miniconda3/pkgs/*.conda 2>/dev/null | wc -l

Why is it so large? Every time conda installs or updates a package, both the download archive and extracted files are kept. If you've been using conda for a year or more, you likely have hundreds of cached packages — many for versions you no longer use.

Clean the Package Cache

# Remove only tarballs (safest — packages stay extracted)
conda clean --tarballs -y

# Remove unused extracted packages
conda clean --packages -y

# Remove both tarballs AND unused packages
conda clean --all -y

# Dry run — see what would be removed without deleting
conda clean --all --dry-run

3. Conda Environments — The Hidden Giants

Each conda environment is a complete, isolated copy of Python and its packages. A single data science environment with PyTorch, TensorFlow, or JAX can easily be 3-8GB. Multiple environments multiply this quickly.

# List all environments with their locations
conda env list

# Size of each environment
du -sh ~/miniconda3/envs/*/ 2>/dev/null | sort -hr
du -sh ~/anaconda3/envs/*/ 2>/dev/null | sort -hr

# Check what packages are in an environment
conda list -n my_env

# Count packages per environment
for env in $(conda env list | grep -v "^#" | awk '{print $1}' | grep -v "^$"); do
    count=$(conda list -n "$env" 2>/dev/null | grep -v "^#" | wc -l)
    echo "$env: $count packages"
done

Remove Unused Environments

# Remove a specific environment
conda env remove -n old_project

# Remove by path (for envs created with --prefix)
conda env remove -p /path/to/env

# Force remove even if active
conda deactivate
conda env remove -n unwanted_env

conda env export -n my_env > my_env_backup.yml

Find Stale Environments

Look for environments you haven't used recently:

# Find envs not accessed in 60+ days (by python binary access time)
for env_dir in ~/miniconda3/envs/*/; do
    env_name=$(basename "$env_dir")
    last_access=$(stat -f "%Sa" -t "%Y-%m-%d" "$env_dir/bin/python" 2>/dev/null || echo "unknown")
    size=$(du -sh "$env_dir" 2>/dev/null | awk '{print $1}')
    echo "$size  $env_name  (last accessed: $last_access)"
done | sort -hr

4. conda clean — The Built-In Cleanup Tool

Conda has a built-in conda clean command with several useful flags:

# The most common cleanup command
conda clean --all -y

# Preview what would be cleaned
conda clean --all --dry-run

# Clean only index cache (useful when having solver issues)
conda clean -i -y

5. Anaconda vs. Miniconda — Why It Matters for Disk Space

The choice between Anaconda and Miniconda has a huge impact on disk usage:

Migrate from Anaconda to Miniconda

# 1. Export your environments first
conda env list
conda env export -n my_project > my_project.yml
conda env export -n ml_env > ml_env.yml

# 2. Remove Anaconda
conda install anaconda-clean -y
anaconda-clean --yes
rm -rf ~/anaconda3/

# 3. Install Miniconda
# Download from https://docs.conda.io/en/latest/miniconda.html
bash Miniconda3-latest-MacOSX-arm64.sh

# 4. Recreate only the environments you need
conda env create -f my_project.yml

6. Index Cache and Metadata

Conda caches channel index data to speed up environment solving. Over time, this metadata accumulates:

# Check index cache size
du -sh ~/.conda/ 2>/dev/null
du -sh ~/miniconda3/pkgs/cache/ 2>/dev/null

# List conda config (channels add to cache size)
conda config --show channels

# Clean index cache
conda clean -i -y

# Remove channel metadata
rm -rf ~/.conda/pkgs_dirs/ 2>/dev/null

If you have many custom channels configured (conda-forge, bioconda, pytorch, nvidia), the index cache can grow to several hundred MB. Cleaning it forces a fresh download on next install, which also fixes solver issues caused by stale metadata.

7. Mamba / Micromamba Caches

If you use Mamba (the faster conda alternative) or Micromamba, they have their own cache locations:

# Mamba uses the same pkgs/ directory as conda
# So conda clean works for mamba too
mamba clean --all -y

# Micromamba has its own root
du -sh ~/micromamba/ 2>/dev/null
du -sh ~/.local/share/mamba/ 2>/dev/null
micromamba clean --all -y

# Check Mambaforge installation
du -sh ~/mambaforge/ 2>/dev/null
du -sh ~/mambaforge/pkgs/ 2>/dev/null

8. Jupyter and Notebook Storage

Conda is often paired with Jupyter, which has its own disk footprint:

# Jupyter data directory
du -sh ~/Library/Jupyter/ 2>/dev/null

# Jupyter kernels (one per environment)
jupyter kernelspec list 2>/dev/null
du -sh ~/Library/Jupyter/kernels/*/ 2>/dev/null

# IPython history and cache
du -sh ~/.ipython/ 2>/dev/null

# Jupyter runtime files
du -sh ~/Library/Jupyter/runtime/ 2>/dev/null

# Notebook checkpoints scattered across projects
find ~ -maxdepth 5 -type d -name ".ipynb_checkpoints" 2>/dev/null | head -20

Clean Jupyter Artifacts

# Remove orphaned kernels (from deleted environments)
jupyter kernelspec list
jupyter kernelspec uninstall old_kernel_name

# Clean notebook checkpoints
find ~/projects -type d -name ".ipynb_checkpoints" -exec rm -rf {} + 2>/dev/null

# Clear IPython history
rm -rf ~/.ipython/profile_default/history.sqlite 2>/dev/null

9. Completely Removing Conda from macOS

If you want to completely remove conda and start fresh (or switch to a different setup):

# 1. Export environments you want to keep
for env in $(conda env list | grep -v "^#" | awk '{print $1}' | grep -v "^$" | grep -v "base"); do
    conda env export -n "$env" > "${env}_backup.yml" 2>/dev/null
    echo "Exported: $env"
done

# 2. Optional: Install anaconda-clean for thorough removal
conda install anaconda-clean -y
anaconda-clean --yes

# 3. Remove the installation directory
rm -rf ~/miniconda3/     # or ~/anaconda3/ or ~/miniforge3/
rm -rf ~/mambaforge/

# 4. Remove conda configuration
rm -rf ~/.conda/
rm -f ~/.condarc

# 5. Remove shell initialization from ~/.zshrc
# Edit ~/.zshrc and remove the conda initialize block:
# >>> conda initialize >>>
# ... (several lines)
# <<< conda initialize <<<

# 6. Remove Jupyter artifacts
rm -rf ~/Library/Jupyter/

# 7. Verify nothing remains
which conda
which python3

10. Automate Conda Cleanup

Add a weekly cleanup to your crontab or create an alias:

# Add to ~/.zshrc for a convenient alias
alias conda-cleanup='conda clean --all -y && echo "Conda cache cleaned!"'

# Or create a full cleanup script
cat << 'EOF' > ~/scripts/conda_cleanup.sh
#!/bin/bash
echo "=== Conda Cleanup ==="
echo "Before:"
du -sh ~/miniconda3/ 2>/dev/null || du -sh ~/anaconda3/ 2>/dev/null

# Clean all caches
conda clean --all -y

# Report stale environments
echo ""
echo "=== Environments (check for unused) ==="
for env_dir in ~/miniconda3/envs/*/; do
    [ -d "$env_dir" ] || continue
    name=$(basename "$env_dir")
    size=$(du -sh "$env_dir" 2>/dev/null | awk '{print $1}')
    echo "  $size  $name"
done | sort -hr

echo ""
echo "After:"
du -sh ~/miniconda3/ 2>/dev/null || du -sh ~/anaconda3/ 2>/dev/null
echo "Done!"
EOF
chmod +x ~/scripts/conda_cleanup.sh

Scheduled Cleanup with cron

# Run conda clean weekly (every Sunday at 3am)
crontab -e
# Add this line:
0 3 * * 0 /Users/$(whoami)/miniconda3/bin/conda clean --all -y >> /tmp/conda_cleanup.log 2>&1

11. Quick Reference Cheatsheet

12. FAQ