oxlint vs ESLint: Should You Switch? 9.8s vs 0.09s

On react-hook-form's 301 lintable files, ESLint 9.39.4 with the repo's own config took a median 9.8 s, and oxlint 1.87.0 with a migrated config took 0.09 s. The catch: only 42 of the 84 ESLint rules moved over natively, and oxlint reported 315 warnings where ESLint reported 562.

Short answer: switch if your config is mostly core, TypeScript and React rules, and keep ESLint (or run both) if you lean on type-aware rules or plugins oxlint cannot run natively. No affiliate links in this post.

What I tested

• Machine: 8 vCPU Intel Xeon / 15 GB RAM Linux cloud box, Node 24.21.0, tested 5 Oct 2026 (about 22:25 to 22:35 IST). The box is shared with other jobs, so I interleaved the tools in each round.
• Repo: react-hook-form at commit 91856c2 (4 Oct 2026), installed from its own lockfile. Its flat config uses typescript-eslint 8.61.1 recommended, eslint-plugin-react, react-hooks, simple-import-sort, eslint-plugin-prettier and eslint-plugin-playwright.
• ESLint: the repo's pinned 9.39.4, run as eslint . with no cache unless noted.
• oxlint: 1.87.0. I generated its config with @oxlint/migrate 1.86.0 twice: once native-only (--js-plugins=false) and once with JS plugins, which loads ESLint plugins oxlint lacks.
• Timing: three runs each, median of wall-clock time for the whole command.
• Seeded test: one extra .tsx file with 15 planted problems, one for each kind of rule in the config, then deleted.
• One surprise: the first time I ran oxlint I kept the generated config one folder up. Ignore patterns and overrides resolve relative to the config file, so oxlint linted 311 files and raised 4 no-require-imports errors ESLint never shows. Keeping .oxlintrc.json in the repo root fixed it.
• Not tested: ESLint 10.12, oxlint's type-aware mode, Vue or Svelte files, editor integration, autofix output.

Is oxlint faster than ESLint?

Setup	Rules	Median	Runs
ESLint 9.39.4, repo config, single thread	84	9.81 s	9.81 / 9.65 / 9.90
ESLint, --concurrency auto	84	6.02 s	7.27 / 6.02 / 5.51
ESLint, prettier/prettier turned off	83	5.66 s	5.66 / 5.22 / 6.28
ESLint, --cache with nothing changed	84	1.10 s	1.10 / 1.22 / 0.92
oxlint, native rules only	42	0.09 s	0.31 / 0.09 / 0.09
oxlint, native + simple-import-sort as a JS plugin	79	0.61 s	0.69 / 0.60 / 0.61
oxlint, native + all JS plugins (incl. Prettier)	80	4.75 s	5.17 / 4.67 / 4.75

Rule counts are as @oxlint/migrate reported them. Native oxlint was about 100x faster than a single-threaded ESLint run. That ratio drops fast once you add JS plugins: running Prettier as a lint rule inside oxlint cost about 4 s, almost the same as running prettier --check 3.8.4 on the same 301 files (4.13 s median). Formatting through the linter is the slow part in both tools.

Two ESLint numbers matter for a fair comparison. Turning on ESLint's built-in multithreading (--concurrency auto) took 9.8 s down to 6.0 s with no other change. And with --cache, an unchanged repo re-linted in 1.1 s, so local re-runs were never the 10 s problem. CI runs without a warm cache are where oxlint pays off.

Does oxlint catch the same problems?

On the real repo, every one of oxlint's 307 unused-variable warnings matched an ESLint warning at the same line and column. Its other 8 warnings (4 react/display-name, 4 Playwright) matched ESLint's per-rule counts. The 247 warnings only ESLint raised were all one pattern in the repo's type-test files: "'actual' is assigned a value but only used as a type." That is a typescript-eslint extension of no-unused-vars, and oxlint's version did not flag it.

The seeded file told the same story:

Planted problem	ESLint	oxlint + JS plugins	oxlint native
Hook called inside if	yes	yes	yes
useEffect missing a dependency	yes	yes	yes
console.log	yes	yes	yes
Type used without import type	yes	yes	yes
Unused variable	yes	yes	yes
Unused type alias	yes	yes	yes
!!x in a condition	yes	yes	yes
// @ts-ignore	yes	yes	yes
var	yes	yes	yes
let never reassigned	yes	yes	yes
Anonymous memo() component	yes	yes	yes
Unsorted imports	yes	yes	no
Formatting off (Prettier)	yes	yes	no
Value only used as a type	yes	no	no
Anonymous default-export component	yes	no	no
Total	15 / 15	13 / 15	11 / 15

The two misses that matter here are both quiet: neither breaks a build, but a team that relies on react/display-name for DevTools names or on the type-only check will lose them without noticing. Check the typescript-eslint 8.71 post if your config uses rules beyond recommended.

What does the migration tool leave out?

@oxlint/migrate read all 84 rules from the flat config. Native-only, it ported 42 and skipped 42: 38 belong to plugins oxlint does not ship (Playwright, Prettier, simple-import-sort), 3 it lists as unsupported (react/jsx-uses-react, react/jsx-uses-vars, react/no-deprecated), and 1 is still a nursery rule (react/require-render-return). With JS plugins on, it ported 80. It also warned that react.version: "detect" and react.pragma settings do not carry over.

JS plugins are still marked alpha in the oxlint docs, and they cannot run type-aware rules yet. If your config uses recommendedTypeChecked, oxlint native alone will not replace it today. The oxlint 1.87 release notes post covers what that version added for React and jsx-a11y.

How do I switch without losing checks?

1. Run npx @oxlint/migrate eslint.config.mjs --details and read the skipped list before you change anything.
2. Save the output as .oxlintrc.json in the repo root, not elsewhere, so ignores and overrides resolve the same way.
3. Move formatting out of the linter. Run Prettier or oxfmt as its own step; the oxfmt vs Prettier test has numbers for that side.
4. Run oxlint first in CI and keep ESLint for only the rules oxlint skipped. eslint-plugin-oxlint exists for exactly this: it turns off the ESLint rules oxlint already covers.
5. Diff the warnings from both tools on your main branch once, the way I did above, before you delete the ESLint config.

Should you switch from ESLint to oxlint?

Situation	My pick
Config is mostly core, typescript-eslint recommended, React and hooks rules	oxlint native, with formatting as a separate step
Heavy use of type-aware rules	Keep ESLint, or run oxlint first and ESLint for the type-aware set
Plugins oxlint does not ship (Playwright, import sorting)	oxlint with JS plugins, or both tools side by side
Vue, Svelte or Angular templates	Keep ESLint for now; JS plugins do not handle custom parsers yet
ESLint is slow only on your laptop	Try --cache and --concurrency auto first

Bottom line: on this repo oxlint went from 9.8 s to 0.09 s and agreed with ESLint on every warning it raised, but it skipped half the rules natively and silently dropped two checks. Who should not switch yet: teams whose lint config is mostly type-aware rules or framework template plugins. Everyone else can get most of the speed by running both and letting oxlint go first.

How this was made: I ran every command above on the ShopperCove test box and kept the raw JSON output; the write-up was drafted with AI help and checked against that output.

Sources

• https://github.com/react-hook-form/react-hook-form
• https://oxc.rs/docs/guide/usage/linter.html
• https://oxc.rs/docs/guide/usage/linter/js-plugins.html
• https://github.com/oxc-project/oxc/releases
• https://www.npmjs.com/package/@oxlint/migrate
• https://eslint.org/docs/latest/use/command-line-interface
• https://eslint.org/blog/2025/08/multithread-linting/
• https://typescript-eslint.io/rules/no-unused-vars/

Related

• https://www.shoppercove.com/blog/oxlint-1-87-react-suggestions-a11y-fixes-october-2026
• https://www.shoppercove.com/blog/eslint-10-12-release-october-2026
• https://www.shoppercove.com/blog/typescript-eslint-8-71-no-unsafe-enum-assignment-october-2026
• https://www.shoppercove.com/blog/oxfmt-0-72-native-markdown-formatter-prettier-parity-october-2026
• https://www.shoppercove.com/blog/biome-2-5-15-type-inference-nursery-rules-october-2026
• https://www.shoppercove.com/blog/vite-plus-1-0-unified-toolchain-october-2026
• https://www.shoppercove.com/blog/vitest-5-should-you-upgrade-october-2026
• https://www.shoppercove.com/blog/playwright-1-63-test-locks-october-2026