lefthook vs husky: Should You Switch? 615ms vs 478ms

On a 40-file prettier pre-commit, husky 9.1.7 plus lint-staged 16.2.3 took a median 478 ms when both used local binaries, and lefthook 1.13.6 took 615 ms (~1.29x). Hook install was close: lefthook 256 ms vs husky 283 ms. With the common npx wrappers the gap stayed in husky's favor (651 ms vs 765 ms).

Short answer: do not switch to lefthook for raw speed on a single prettier hook. Stay on husky plus lint-staged if that stack already works. Move to lefthook when you want a YAML-native Go runner, monorepo-friendly parallel commands, or to drop the lint-staged Node layer. No affiliate links in this post.

What I tested

• Machine: 8 vCPU Intel Xeon / 15 GB RAM Linux cloud box, tested 5 Oct 2026 (about 23:20 to 23:25 IST). Shared box; variants interleaved each round.
• Fixture: two git repos with the same 40 unformatted .js modules staged for commit. Task: run prettier 3.6.2 --write on those staged files.
• Tools: husky 9.1.7 with lint-staged 16.2.3; lefthook 1.13.6. Node at the time of the hook runs was v22.14.0 via the Volta default on the box.
• Timing: one warmup each, then five interleaved rounds. Wall clock via process.hrtime around the hook command (not a full git commit, so empty-commit noise is out of the number).
• Findings: local-bin medians 477.5 ms (husky/lint-staged) vs 614.5 ms (lefthook calling node_modules/.bin/prettier). Install medians 282.5 ms (husky) vs 255.9 ms (lefthook). Bare prettier without any hook framework: 464 ms.
• One surprise: lefthook's Go binary did not win this single-command case. Most of the wall time is prettier plus process startup; an extra runner layer still shows up at this size.
• Not tested: many parallel lefthook commands, monorepo per-package hooks, eslint on TypeScript, Windows/macOS, pre-push, commit-msg.

Is lefthook faster than husky on a prettier pre-commit?

Variant	Median wall	Runs (ms)
husky path: local lint-staged --allow-empty	478 ms	478 / 459 / 474 / 478 / 488
lefthook path: local lefthook + local prettier	615 ms	614 / 617 / 615 / 621 / 610
husky path: npx lint-staged --allow-empty	651 ms	628 / 651 / 630 / 694 / 689
lefthook path: npx lefthook + npx prettier	765 ms	758 / 750 / 765 / 787 / 884
Bare npx prettier --write src/*.js (no hook)	464 ms	457 / 468 / 468 / 464 / 463

No. On this 40-file fixture husky plus lint-staged was faster in both the local-binary and npx setups. Lefthook still beats a cold full Node toolchain story when you configure many parallel commands in YAML; that is not what I measured here. If your pre-commit is mostly format-on-save style work, see the dprint vs Prettier post and the Biome vs ESLint/Prettier post for the formatter choice underneath the hook.

How long does hook install take?

Install	Median wall	Runs (ms)
npx husky then write .husky/pre-commit	283 ms	277 / 283 / 344 / 348 / 265
npx lefthook install	256 ms	255 / 305 / 265 / 256 / 254

Almost a wash. Lefthook was about 27 ms faster to wire the git hook on this box. Neither install time should decide a migration.

Should you switch from husky to lefthook?

Situation	My pick
Single prettier or eslint pre-commit, already on lint-staged	Stay on husky; it was faster here
You want YAML config and a Go runner checked into the repo	lefthook
Monorepo with many parallel hook commands	lefthook (not measured; that is the product pitch)
Team is happy and hooks finish under a second	Do not switch this quarter
You are choosing a linter/formatter first	Fix that before the hook runner; see oxlint vs ESLint and knip vs depcheck​

Bottom line: husky plus lint-staged won the 40-file prettier hook (478 ms vs 615 ms local binaries; 651 ms vs 765 ms with npx). lefthook install was slightly quicker (256 ms vs 283 ms). Who should not switch yet: repos whose only pain is "prettier on staged files" and who already ship husky. Who should try lefthook: teams that outgrew a one-line lint-staged config and want parallel, YAML-defined hooks.

How this was made: I built two identical 40-file git fixtures, ran interleaved hook and install timings on the ShopperCove test box, and kept the JSON; the write-up was drafted with AI help and checked against that output.

Sources

• 01github.com/typicode/husky
• 02github.com/lint-staged/lint-staged
• 03github.com/evilmartians/lefthook
• 04npmjs.com/package/husky
• 05npmjs.com/package/lefthook
• 06github.com/prettier/prettier

Related

• 01shoppercove.com/blog/turbo-vs-nx
• 02shoppercove.com/blog/madge-vs-dependency-cruiser
• 03shoppercove.com/blog/pnpm-vs-npm-vs-bun
• 04shoppercove.com/blog/c8-vs-nyc