Measured on 2026-10-08 with zerobrew 0.3.5 (development release) and Homebrew 7.0.8 on macOS 26.6.2, MacBook Pro (M3 Pro, 18 GB RAM), ~318 Mbit/s download bandwidth. Homebrew started with nothing installed.
- **Cold**: the package and all of its dependencies uninstalled, empty download cache.
- **Warm**: the package and all of its dependencies uninstalled, downloads from the cold run still cached.
- The median package was 5.3x faster cold and 56x faster warm. Per package, cold ranged from 1.6x (go) to 20x (ca-certificates) and warm from 18x (go) to 252x (ca-certificates). 24 of the 100 packages installed 100x faster or more warm, which is what the asterisk on the tagline refers to.
- Cold installs are bound by the link. The same run on a ~69 Mbit/s connection ([results/2026-10-07](results/2026-10-07/benchmark.md)) came out at 3.3x cold and 69x warm. Big bottles like go and llvm spend nearly all of their cold time downloading, so both tools land close together there.
Full results
| Package | Homebrew (cold) | ZB (cold) | Cold speedup | Homebrew (warm) | ZB (warm) | Warm speedup |
|---------|-----------------|-----------|--------------|-----------------|-----------|--------------|
| **Overall (100 packages)** | 775.60s | 117.43s | **6.60x** | 637.50s | 9.33s | **68.31x** |
| ca-certificates | 7.65s | 376ms | 20.35x | 6.55s | 26ms | 252.12x |
| openssl@3 | 11.02s | 1.30s | 8.46x | 9.18s | 64ms | 143.47x |
| xz | 2.39s | 454ms | 5.26x | 1.46s | 29ms | 50.31x |
| sqlite | 2.99s | 591ms | 5.06x | 2.04s | 27ms | 75.67x |
| readline | 2.48s | 494ms | 5.02x | 1.46s | 26ms | 56.08x |
| icu4c@78 | 3.32s | 1.57s | 2.11x | 1.96s | 33ms | 59.42x |
| python@3.14 | 16.99s | 1.97s | 8.62x | 14.37s | 207ms | 69.41x |
| awscli | 27.22s | 3.37s | 8.08x | 25.56s | 483ms | 52.92x |
| node | 29.31s | 3.16s | 9.28x | 28.24s | 232ms | 121.72x |
| harfbuzz | 29.34s | 2.20s | 13.34x | 22.35s | 364ms | 61.41x |
| ncurses | 3.13s | 1.04s | 3.01x | 2.04s | 69ms | 29.59x |
| gh | 2.60s | 903ms | 2.87x | 1.35s | 38ms | 35.58x |
| pcre2 | 2.58s | 631ms | 4.08x | 1.57s | 35ms | 44.94x |
| libpng | 2.49s | 854ms | 2.92x | 1.50s | 27ms | 55.70x |
| zstd | 4.12s | 672ms | 6.13x | 2.74s | 34ms | 80.50x |
| glib | 7.02s | 1.72s | 4.08x | 5.24s | 94ms | 55.76x |
| lz4 | 2.89s | 635ms | 4.54x | 1.48s | 27ms | 54.74x |
| gettext | 4.72s | 1.30s | 3.61x | 3.25s | 64ms | 50.81x |
| libngtcp2 | 10.79s | 1.40s | 7.73x | 9.20s | 66ms | 139.36x |
| libnghttp3 | 2.38s | 436ms | 5.47x | 1.47s | 27ms | 54.44x |
| pkgconf | 2.35s | 469ms | 5.02x | 1.47s | 27ms | 54.59x |
| libunistring | 2.57s | 981ms | 2.62x | 1.49s | 26ms | 57.12x |
| mpdecimal | 2.53s | 705ms | 3.59x | 1.47s | 27ms | 54.33x |
| brotli | 2.65s | 776ms | 3.41x | 1.48s | 27ms | 54.67x |
| jpeg-turbo | 2.55s | 889ms | 2.87x | 1.48s | 27ms | 54.78x |
| xorgproto | 2.21s | 484ms | 4.57x | 1.21s | 29ms | 41.59x |
| ffmpeg | 18.57s | 1.76s | 10.54x | 17.62s | 164ms | 107.41x |
| cmake | 4.04s | 1.66s | 2.44x | 1.73s | 80ms | 21.64x |
| libnghttp2 | 2.63s | 635ms | 4.14x | 1.46s | 26ms | 56.31x |
| go | 5.43s | 3.36s | 1.61x | 3.40s | 190ms | 17.92x |
| uv | 2.90s | 1.37s | 2.12x | 1.27s | 28ms | 45.43x |
| gmp | 2.62s | 543ms | 4.83x | 1.47s | 26ms | 56.42x |
| libtiff | 7.08s | 821ms | 8.62x | 6.11s | 53ms | 115.19x |
| fontconfig | 9.35s | 982ms | 9.52x | 8.13s | 77ms | 105.61x |
| python@3.13 | 14.28s | 1.76s | 8.12x | 13.15s | 186ms | 70.71x |
| git | 6.46s | 1.67s | 3.88x | 4.98s | 101ms | 49.33x |
| little-cms2 | 7.85s | 792ms | 9.91x | 6.83s | 55ms | 124.16x |
| dav1d | 2.49s | 544ms | 4.58x | 1.47s | 26ms | 56.42x |
| openexr | 9.95s | 712ms | 13.98x | 8.83s | 59ms | 149.59x |
| c-ares | 2.51s | 513ms | 4.90x | 1.51s | 32ms | 47.34x |
| tesseract | 34.63s | 2.16s | 16.01x | 32.37s | 476ms | 68.01x |
| p11-kit | 9.26s | 996ms | 9.30x | 7.95s | 32ms | 248.53x |
| imagemagick | 14.51s | 1.21s | 12.00x | 13.20s | 82ms | 160.95x |
| zlib | 2.36s | 430ms | 5.48x | 1.46s | 25ms | 58.44x |
| libx11 | 5.70s | 932ms | 6.11x | 4.71s | 174ms | 27.05x |
| freetype | 3.51s | 585ms | 5.99x | 2.21s | 32ms | 69.22x |
| protobuf | 4.06s | 725ms | 5.60x | 3.09s | 57ms | 54.21x |
| gnupg | 21.97s | 1.36s | 16.16x | 20.63s | 192ms | 107.44x |
| openjph | 7.88s | 697ms | 11.31x | 6.80s | 54ms | 126.02x |
| libtasn1 | 2.30s | 461ms | 4.99x | 1.49s | 28ms | 53.25x |
| ruby | 16.01s | 3.83s | 4.18x | 14.56s | 801ms | 18.18x |
| gnutls | 14.93s | 1.19s | 12.52x | 13.93s | 134ms | 103.96x |
| expat | 2.40s | 442ms | 5.42x | 1.48s | 26ms | 56.88x |
| libsodium | 2.44s | 884ms | 2.76x | 1.48s | 27ms | 54.96x |
| simdjson | 2.52s | 619ms | 4.08x | 1.49s | 26ms | 57.12x |
| gemini-cli | 29.98s | 2.23s | 13.43x | 27.64s | 235ms | 117.60x |
| libarchive | 5.02s | 594ms | 8.45x | 4.04s | 35ms | 115.31x |
| pyenv | 12.61s | 1.39s | 9.06x | 11.44s | 135ms | 84.76x |
| pixman | 2.44s | 487ms | 5.01x | 1.49s | 26ms | 57.19x |
| curl | 16.37s | 1.47s | 11.12x | 15.20s | 173ms | 87.88x |
| opus | 2.37s | 497ms | 4.76x | 1.47s | 27ms | 54.59x |
| unbound | 12.24s | 1.35s | 9.07x | 11.03s | 113ms | 97.58x |
| cairo | 19.28s | 1.48s | 12.99x | 17.74s | 370ms | 47.95x |
| pango | 24.48s | 1.69s | 14.46x | 23.06s | 411ms | 56.10x |
| leptonica | 9.48s | 914ms | 10.37x | 8.39s | 66ms | 127.18x |
| libxcb | 4.88s | 887ms | 5.50x | 3.89s | 132ms | 29.45x |
| jpeg-xl | 12.96s | 997ms | 13.00x | 11.92s | 70ms | 170.31x |
| coreutils | 3.31s | 760ms | 4.35x | 2.24s | 38ms | 58.97x |
| certifi | 7.81s | 461ms | 16.95x | 6.98s | 31ms | 225.06x |
| krb5 | 10.49s | 1.34s | 7.85x | 9.28s | 68ms | 136.44x |
| docker | 2.26s | 734ms | 3.08x | 1.21s | 29ms | 41.83x |
| libheif | 10.88s | 961ms | 11.32x | 9.59s | 63ms | 152.24x |
| webp | 4.31s | 578ms | 7.46x | 3.37s | 35ms | 96.40x |
| libxext | 6.45s | 982ms | 6.57x | 5.42s | 182ms | 29.80x |
| libxau | 2.74s | 512ms | 5.35x | 1.85s | 33ms | 56.06x |
| gcc | 12.23s | 3.55s | 3.45x | 8.58s | 67ms | 128.00x |
| bzip2 | 2.09s | 368ms | 5.69x | 1.14s | 26ms | 43.65x |
| libxdmcp | 2.74s | 540ms | 5.08x | 1.86s | 33ms | 56.42x |
| abseil | 3.05s | 800ms | 3.82x | 1.90s | 50ms | 38.04x |
| xcbeautify | 2.22s | 481ms | 4.62x | 1.15s | 26ms | 44.35x |
| libuv | 2.51s | 482ms | 5.20x | 1.47s | 26ms | 56.65x |
| giflib | 2.31s | 673ms | 3.44x | 1.46s | 26ms | 56.12x |
| utf8proc | 2.36s | 452ms | 5.23x | 1.46s | 27ms | 54.15x |
| libxrender | 6.35s | 976ms | 6.51x | 5.42s | 179ms | 30.25x |
| m4 | 2.00s | 431ms | 4.65x | 1.13s | 25ms | 45.28x |
| graphite2 | 2.47s | 428ms | 5.78x | 1.46s | 26ms | 56.23x |
| openjdk | 32.09s | 5.17s | 6.21x | 26.86s | 448ms | 59.96x |
| uvwasi | 2.95s | 483ms | 6.10x | 2.11s | 29ms | 72.66x |
| libffi | 2.56s | 575ms | 4.46x | 1.43s | 26ms | 55.08x |
| libdeflate | 2.44s | 458ms | 5.33x | 1.46s | 26ms | 56.12x |
| llvm | 18.22s | 9.14s | 1.99x | 9.66s | 183ms | 52.79x |
| aom | 3.25s | 699ms | 4.65x | 2.16s | 32ms | 67.50x |
| lzo | 2.38s | 449ms | 5.31x | 1.46s | 26ms | 56.00x |
| libevent | 10.66s | 1.33s | 7.99x | 9.19s | 71ms | 129.42x |
| libgpg-error | 5.06s | 1.11s | 4.56x | 3.84s | 70ms | 54.81x |
| libidn2 | 4.97s | 1.09s | 4.56x | 3.87s | 69ms | 56.12x |
| berkeley-db@5 | 4.05s | 1.96s | 2.06x | 2.60s | 56ms | 46.45x |
| deno | 10.57s | 1.46s | 7.23x | 9.24s | 57ms | 162.09x |
| libedit | 2.35s | 442ms | 5.33x | 1.45s | 26ms | 55.65x |
| oniguruma | 2.37s | 508ms | 4.67x | 1.46s | 26ms | 56.19x |
To reproduce, run `just bench --full results/` on a machine with nothing installed in Homebrew. It resets zerobrew, uninstalls everything it installed in Homebrew, and writes a table to `results/benchmark.md`.
A smaller version runs in CI: the [parity workflow](https://github.com/zerobrewhq/zerobrew/actions/workflows/parity.yml) installs a fixed set of formulae with both tools on the same runner and compares the resulting prefixes byte for byte, and the [timing workflow](https://github.com/zerobrewhq/zerobrew/actions/workflows/timing.yml) times those formulae plus node and python@3.14, cold and warm, with both tools and with the last zerobrew release.
The nightly timing run fails if zerobrew is less than 1.5x faster than Homebrew cold or 3x warm on any of them, or more than 20% slower than the last release overall.
## A note on how we are achieving these speeds
WRT Homebrew, I want it to be clear that both `brew` and `zerobrew` install the same bottles from *Homebrew's build farm*.
The difference is what happens after the download: Homebrew runs Ruby to evaluate the formula, unpacks, rewrites paths with `install_name_tool`, and re-signs each binary one at a time, while zerobrew **does the relocation in-process** and links from a content-addressed store, so it pays for the download once and almost nothing after.
**I want it to be clear that we are basically nothing without Homebrew's bottle build farm**. Every package zerobrew installs is one they built, tested and published. We don't compile anything, we don't maintain formulae, and none of these numbers exist without that work. The speed comes from how the artifacts are installed, not from what's in them.
This is a good read, if you're interested in what zerobrew is and isn't: [Standing on the shoulders of Homebrew](https://nesbitt.io/2026/04/14/standing-on-the-shoulders-of-homebrew.html) by Andrew Nesbitt. It makes the point better than I can.
**nanobrew:**
[nanobrew’s](https://github.com/justrach/nanobrew) published numbers time a “warm install” of a package that is already installed, which is an early exit, not an install, and compare against zerobrew 0.1.0.
Every number above is a real install from an uninstalled state, with the script and the full log in this repo, so you can run it yourself.
## Relationship with Homebrew
zerobrew is more of a performance-optimized client for the Homebrew ecosystem. We rely on:
- Homebrew's formula definitions (homebrew-core)
- Homebrew's pre-built bottles when available
- Homebrew's package metadata and infrastructure
Our innovations focus on:
- Content-addressable storage for deduplication
- APFS clonefiles for zero-overhead copying
- Source build fallback using Homebrew's Ruby DSL
zerobrew is experimental. We recommend running it alongside Homebrew rather than as a replacement, and do _not_
recommend purging homebrew and replacing it with zerobrew unless you are absolutely sure about the implications of
doing so.
## Project status
- **Status:** Experimental, but quite useful. I ([@cachebag](https://github.com/cachebag)) daily drive it myself.
- **Feedback:** If you hit incompatibilities, please open an issue or PR.
- **License:** Dual-licensed under [Apache 2.0](./LICENSE-APACHE.md) OR [MIT](./LICENSE-MIT.md), at your choice.
