Install
# One-line install (auto-detects GPU, pinned to this release)
curl -fsSL https://raw.githubusercontent.com/utensils/mold/v0.20.2/install.sh | MOLD_VERSION=v0.20.2 sh
# From crates.io (CPU-only, no GPU acceleration)
cargo install mold-ai
# From source with GPU support
cargo install --git https://github.com/utensils/mold --tag v0.20.2 --features metal,preview,discord,expand,tui,webp,mp4,metrics,mdns # macOS
cargo install --git https://github.com/utensils/mold --tag v0.20.2 --features cuda,preview,discord,expand,tui,webp,mp4,metrics,mdns # Linux
# Nix
nix run github:utensils/mold/v0.20.2Download pre-built binaries
| Platform | GPU | File |
|---|---|---|
| macOS Apple Silicon | Metal | mold-aarch64-apple-darwin.tar.gz |
| Linux x86_64 | CUDA — RTX 40-series (Ada) | mold-x86_64-unknown-linux-gnu-cuda-sm89.tar.gz |
| Linux x86_64 | CUDA — RTX 50-series (Blackwell) | mold-x86_64-unknown-linux-gnu-cuda-sm120.tar.gz |
Desktop app (macOS)
Mold-macos-arm64.dmg (also attached with its versioned name) — signed, notarized, and stapled. Drag Mold to Applications; no quarantine dance needed. Stable link to the newest build: https://github.com/utensils/mold/releases/latest/download/Mold-macos-arm64.dmg
What's Changed
- fix(web): fake installed models list, and unusable phone navigation by @jamesbrink in #472
- docs(changelog): record the web fixes released as 0.20.2 by @jamesbrink in #475
- chore: release v0.20.2 by @release-plz-mold[bot] in #474
Full Changelog: v0.20.1...v0.20.2