Rspack vs Webpack — Is the Rust Bundler Ready to Replace?

If your webpack build is slow, switch to Rspack. Dev startup is 5–18× faster, HMR is 31× faster at scale, and production builds are 5–9× faster — all with the same webpack config and near-identical bundle output. Migration takes 1–2 days for a medium-sized app.

Who this is for

Teams running webpack 5 projects who are tired of waiting for builds. If you are starting a new project with no webpack history, read the verdict section first — Vite 8 may be the better default.

What we tested

Rspack v2.0.6 (released 2026-06-02) vs webpack 5.107.2. Both configured with SWC (not Babel) for apples-to-apples comparison. Benchmark data from rstackjs/build-tools-performance, GitHub Actions run 2026-06-01. App scales: react-1k (1,000 modules), react-5k, react-10k.

The numbers

The headline metric is HMR at scale: 31× faster at 5,000 modules — 105 ms vs 3,314 ms. For a developer who triggers dozens of saves per hour, that difference is the difference between a tight feedback loop and context-switching out to a browser tab.

Dev server startup (no cache)

App scale	Rspack 2.0.6	webpack 5.107	Speedup
1,000 modules	1,566 ms	9,135 ms	5.8×
5,000 modules	901 ms	14,189 ms	15.7×
10,000 modules	1,476 ms	26,984 ms	18.3×

Hot module replacement

App scale	Rspack 2.0.6	webpack 5.107	Speedup
1,000 modules	163 ms	736 ms	4.5×
5,000 modules	105 ms	3,314 ms	31.6×
10,000 modules	156 ms	3,386 ms	21.7×

Production build (no cache)

App scale	Rspack 2.0.6	webpack 5.107	Speedup
1,000 modules	1,032 ms	7,730 ms	7.5×
5,000 modules	2,014 ms	14,071 ms	7.0×
10,000 modules	5,219 ms	45,436 ms	8.7×

Memory usage (dev server RSS)

App scale	Rspack 2.0.6	webpack 5.107	Rspack advantage
1,000 modules	347 MB	822 MB	2.4× less
5,000 modules	283 MB	1,818 MB	6.4× less
10,000 modules	355 MB	1,923 MB	5.4× less

Bundle output

Bundle sizes are essentially identical. At react-10k: Rspack outputs 5,861 kB vs webpack’s 5,934 kB — a 1.2% difference. Gzipped sizes are within 0.1–1% across all scenarios. You do not trade output quality for build speed.

One note on the cached dev startup: webpack 5.107.2 at react-10k is slower with cache (27,569 ms) than without (26,984 ms), suggesting cache invalidation overhead that compounds at scale.

Compatibility

Rspack’s design goal is drop-in replacement for webpack 5. In practice, >80% of the top-50 webpack plugins work natively or with a maintained alternative.

Works without changes

28 widely-used plugins are fully compatible, including html-webpack-plugin, webpack-bundle-analyzer, terser-webpack-plugin, compression-webpack-plugin, @sentry/webpack-plugin (v1.20.1+), and @loadable/webpack-plugin. Almost all loaders work: babel-loader, css-loader, sass-loader, less-loader, vue-loader, svelte-loader.

Built-in replacements (identical config)

Some webpack plugins ship as built-ins in Rspack — drop in, no config changes:

webpack	Rspack built-in
`copy-webpack-plugin`	`rspack.CopyRspackPlugin`
`mini-css-extract-plugin`	`rspack.CssExtractRspackPlugin`
`tsconfig-paths-webpack-plugin`	`resolve.tsConfig` option
`circular-dependency-plugin`	`CircularDependencyRspackPlugin`

Drop-in alternatives

Three common plugins require a maintained Rspack-specific package:

webpack	Rspack alternative
`eslint-webpack-plugin`	`eslint-rspack-plugin`
`fork-ts-checker-webpack-plugin`	`ts-checker-rspack-plugin`
`@pmmmwh/react-refresh-webpack-plugin`	`@rspack/plugin-react-refresh`

Not yet supported (8 plugins)

@cypress/webpack-preprocessor, critters-webpack-plugin, git-revision-webpack-plugin, webpack-remove-empty-scripts, and four others. See the full compatibility list. If your project depends on any of these, migration requires either a workaround or waiting for coverage.

Module Federation

Rspack ships first-class Module Federation support at two levels. MF 1.5 is a drop-in for existing federated webpack apps — same API. MF 2.0 (via @module-federation/enhanced) adds dynamic type hints, a Chrome DevTools extension, runtime plugins, and preloading. Interop between webpack hosts and Rspack remotes is supported, which makes incremental monorepo migration viable.

Migration

The mechanical steps are straightforward:

# 1. Swap packages
npm remove webpack webpack-cli webpack-dev-server
npm install -D @rspack/core @rspack/cli @rspack/dev-server

# 2. Rename config
mv webpack.config.js rspack.config.js

# 3. Update scripts in package.json
#   "build": "webpack"       →  "build": "rspack build"
#   "dev": "webpack serve"   →  "dev": "rspack serve"

# 4. Replace mini-css-extract-plugin with the built-in
#   Remove the import; use rspack.CssExtractRspackPlugin instead

After that, swap babel-loader for the built-in builtin:swc-loader to pick up an additional 2–4× speed gain:

// rspack.config.js
{
  test: /\.[jt]sx?$/,
  use: [{
    loader: 'builtin:swc-loader',
    options: { jsc: { parser: { syntax: 'typescript', tsx: true } } }
  }]
}

Watch for these friction points (all hit by Mews in their real-world migration):

Dynamic imports with multiple string interpolations: SWC doesn’t support import(\${dir}/${file}`)` with multiple interpolations. Requires refactoring the call site.
Styled Components SWC plugin: swc_core version alignment between Rspack and the styled-components SWC plugin needs care — pin both and verify on upgrade.
Browserslist: the Rust implementation requires BROWSERSLIST_DANGEROUS_EXTEND=true as an env var if you’re extending configurations across packages.
Custom Babel transforms: Babel plugins have no direct Rspack equivalent. Rewrite as SWC plugins or standalone transforms.

Recommended rollout: pilot one non-critical app, validate outputs match (the bundle size numbers above confirm they should), then roll out to the rest. ByteDance runs Rspack internally across TikTok, Douyin, Lark, and Coze — the rollout path is well-worn.

For monorepos, Nx ships native Rspack support via @nx/rspack. It handles the per-app config wiring and lets you migrate one app at a time without touching the root build setup.

Verdict

Migrate if: your webpack builds are slow and none of your plugins are on the unsupported list. That covers most teams. The benchmarks are confirmed by independent runs and large-scale adopters including Microsoft, Amazon, Alibaba, and Discord. Bundle output is byte-for-byte comparable.

Wait if: your project depends on @cypress/webpack-preprocessor, critters-webpack-plugin, or another unsupported plugin and there is no viable alternative. Also wait if you have complex Babel transforms you cannot rewrite.

Skip webpack entirely if: you are starting a new project with no webpack history. Rspack is the right choice if you need webpack’s plugin surface. If you do not, Vite 8 (with Rolldown+Oxc) is faster at cold build on the popular-libs benchmark — 1,023 ms vs 1,668 ms — and produces slightly smaller bundles.

The Rust bundler is production-ready. As of Rspack 2.0 (April 2026), the question is not whether it works — 5 million weekly downloads and enterprise deployment at scale answer that. The question is whether your specific plugin list fits. For most projects, it does.

Summary

Dimension	webpack 5.107	Rspack 2.0.6
Cold dev start (10k modules)	~27s	~1.5s (18×)
HMR (5k modules)	~3.3s	~105ms (31×)
Prod build (10k modules)	~45s	~5.2s (8.7×)
Dev memory (10k modules)	~1.9 GB	~355 MB (5.4×)
Bundle output size	✅	✅ (≈ identical)
Config compatibility	Native	~85% drop-in
Top-50 plugin support	Native	>80% (native/alt)
Module Federation	MF 1.x	MF 1.5 + 2.0
npm downloads/week	>30M	~5M

Caveats

Rspack v2.0.6 benchmarks were run against Rspack 2.0.5 (one patch behind) — the benchmark repo had not updated when the data was collected on 2026-06-01. No functional differences between 2.0.5 and 2.0.6 affect the numbers.

In December 2024, the Rspack npm packages were briefly compromised with crypto-mining malware (The Hacker News). The incident was resolved within hours. All v1.1+ releases are clean. For security-sensitive deployments, pin versions and verify checksums in CI.

npm download comparisons between webpack (>30M/week) and Rspack (~5M/week) reflect webpack’s 15-year install base, not current adoption intent. The more meaningful signal: Rspack went from 100k/week at v1.0 (October 2024) to 5M/week at v2.0 (April 2026) — 50× growth in 18 months.

This article contains no affiliate links.