JavaScript Bundlers in Late 2026: Vite 8, Turbopack, Rspack, webpack, esbuild and Bun
Most developers don't choose a bundler any more. You choose Next.js, or Vite, or a Vite-based framework, and the bundler comes with it. That's mostly fine, until a build gets slow, a plugin stops working after an upgrade, or someone asks whether the webpack config from 2021 has to go. Then you need to know what's actually under the hood in each tool, and which of them are still moving.
Here's where things stand as of November 2026, with versions and dates taken from each project's own release posts.
Vite 8: one bundler instead of two
For years Vite used esbuild in development and Rollup for production builds. Two tools meant two sets of behaviour, and occasionally a bug that showed up only in the production bundle. Vite 8, released on 12 March 2026, replaced both with Rolldown, a Rust bundler, paired with the Oxc compiler.
What that means in practice:
- Most configs keep working. Vite 8 ships a compatibility layer that maps existing
esbuildandbuild.rollupOptionssettings to their Rolldown and Oxc equivalents. For large projects the Vite team suggested a two-step path: switch to therolldown-vitepreview package on Vite 7 first, then upgrade. - Builds are faster, by how much depends on you. The announcement says "up to 10-30x faster builds", and Rolldown's site attributes the 10-30x figure to its own benchmarks against Rollup. The more useful numbers are the user reports in the post: Linear's production build went from 46s to 6s, and other companies reported reductions between 38% and 64%. Expect something in that range, not a guaranteed multiple.
- Node.js 20.19+ or 22.12+ is required, same as Vite 7.
Vite 8.1 (23 June 2026) added an experimental Bundled Dev Mode, previously called Full Bundle Mode: the dev server serves bundled files instead of one request per module, aimed at large apps where unbundled dev startup gets slow. You turn it on with --experimental-bundle or experimental.bundledDev: true. The release notes say third-party plugins may not work with it yet, so treat it as something to try on a branch. 8.1 also lets you try Lightning CSS as the CSS transformer (css.transformer: 'lightningcss'), which the team is considering as the default in the next major.
Rolldown itself is on the 1.x line; v1.3.0 shipped on 5 November 2026.
Vitest moves with Vite: Vitest 5 requires Vite 6.4.0 or later and Node.js 22.12.0 or later, so check both before upgrading your test runner. Our Jest vs Vitest comparison covers the test-runner side.
Turbopack: the default, but only in Next.js
Next.js 16 (21 October 2025) made Turbopack stable for both next dev and next build, and the default bundler for all apps. Vercel's own figures in that post are "2–5× faster production builds" and "up to 10× faster Fast Refresh". Since then, the Next.js 16.x releases have kept adding to it: 16.3 brought a persistent file cache for builds, and 16.4 (6 October 2026) is the current minor.
Two things to know before upgrading an older app, both from the Turbopack docs:
- webpack plugins don't run. Turbopack supports webpack loaders through
turbopack.rules, but not plugins, and awebpack()function innext.config.jsis ignored. If you depend on a plugin, keep webpack withnext build --webpackuntil there's an alternative. - Small output differences. CSS Modules follow JS import order more strictly, the legacy
~prefix for Sass imports fromnode_modulesisn't supported, and Lightning CSS rounds numbers to 5 decimal places where webpack used 10. These show up as subtle visual diffs, not build errors.
Turbopack doesn't type-check (run tsc separately), and it isn't available as a general-purpose bundler outside Next.js. If you're on Next.js, use it; there isn't a separate decision to make. Our Next.js 16 upgrade guide covers the rest of the upgrade.
Rspack and Rsbuild: the way out of a big webpack config
Rspack is a Rust reimplementation that keeps webpack's configuration model and plugin API, which makes it the realistic option when a large webpack config is too much work to rewrite for Vite. Rspack 2.0 shipped on 22 April 2026, with Rsbuild 2.0, the higher-level build tool built on Rspack, released alongside. The biggest breaking change: core packages such as @rspack/core are now pure ESM, with the CommonJS builds removed. The team says webpack ecosystem compatibility "will remain an important goal" through 2.x. The latest minor is 2.3, from 29 October 2026.
If you migrate, expect most loaders and many plugins to work, then test the ones that touch webpack internals. "Compatible API" doesn't mean every plugin behaves identically.
webpack: maintained, volunteer-run, heading for 6
webpack is still shipping: 5.112 came out on 13 October 2026, with a minor release roughly every month since April (the webpack blog lists 5.106 through 5.112). The 2026 roadmap is open about its position: the project is "maintained primarily by volunteers" with no full-time team, and raising donations is a priority. The work it describes is aimed at webpack 6: a universal target that outputs pure ESM for any runtime (not finished yet), native CSS support graduating from experiments.css so mini-css-extract-plugin is no longer needed, moving loader-runner into core, and evaluating Rspack's lazy barrel optimization.
My read: a working webpack 5 setup isn't an emergency. If you're starting something new, I wouldn't pick it. If an existing setup hurts, Rspack is the low-friction move and Vite is the bigger, cleaner one; our webpack vs Vite comparison and the Create React App migration guide cover that path.
esbuild and Bun: bundlers you call directly
esbuild still isn't 1.0: the latest release on GitHub is v0.28.2. It's no longer inside Vite 8's pipeline, but it remains a good choice for bundling a library, a CLI or a serverless function where you want one fast command and no framework.
Bun's bundler (docs) is built into the runtime:
bun build ./index.tsx --outdir ./build
bun build ./cli.tsx --outfile mycli --compile # single executable
Targets are browser (default), bun and node. The --compile flag produces a standalone executable with the Bun runtime embedded, which is its most distinctive feature. The docs list limits worth knowing: no type checking unless you pass --check, no down-levelling of modern syntax, and CJS and IIFE output are still experimental. For a CLI or a server you already run on Bun, it's convenient; for a browser app with a plugin ecosystem, a framework-backed bundler is the safer pick. See Bun vs Node.js if the runtime is the real question.
Who's behind Vite now
On 4 June 2026, VoidZero, the company behind Vite, Vitest, Rolldown, Oxc and Vite+, joined Cloudflare. The Vite team's post says Vite "remains open source under the MIT license" and "remains vendor-agnostic", is still led by a team with members from several organizations, and that the same terms cover Vitest, Rolldown, Oxc and Vite+. Cloudflare also announced "a new $1M open source fund for the Vite ecosystem."
So the two most-used dev bundlers are now each backed by a platform company: Turbopack by Vercel, Vite's toolchain by Cloudflare. Neither has changed how you configure your project, but it's worth watching where integrations land first.
Picking one
| Situation | Use | Why |
|---|---|---|
| Next.js app | Turbopack (default) | Nothing to choose; --webpack only if a plugin forces it |
| New SPA, React/Vue/Svelte | Vite 8 | One bundler (Rolldown) for dev and prod |
| Big webpack 5 app, rewrite not on the table | Rspack / Rsbuild 2 | Same config model, mostly the same plugins |
| Library or Node CLI | esbuild or Rolldown directly | One command, no framework |
| Service already on Bun | bun build | Built in; --compile for single binaries |
| Working webpack 5 setup, no pain | Leave it | It's maintained; migrate when there's a reason |
Whichever you land on, measure before and after on your own repo: time a cold production build, a warm one, and a dev-server start. Vendor multiples are a ceiling, not a forecast. And if smaller output is the actual goal, the bundler matters less than what you ship: code splitting and lazy loading usually moves the numbers more.
Written by AIDToolStack Editorial
Researched from official documentation, specifications and vendor pages, listed under Sources below. Last checked November 2026.
Sources
- vite.dev/blog/announcing-vite8
- vite.dev/blog/announcing-vite8-1
- vite.dev/blog/cloudflare-supports-vite
- vite.dev/blog
- rolldown.rs/
- github.com/rolldown/rolldown/releases
- vitest.dev/blog/vitest-5
- nextjs.org/blog/next-16
- nextjs.org/blog
- nextjs.org/docs/app/api-reference/turbopack
- rspack.rs/blog/announcing-2-0
- rspack.rs/blog/
- webpack.js.org/blog/
- webpack.js.org/blog/2026-02-04-roadmap-2026/
- github.com/evanw/esbuild/releases
- bun.com/docs/bundler
Explore on AIDToolStack
Discover tools and servers featured in our directory.
AI Tools
MCP Servers
Sponsor Our Newsletter
Reach thousands of developers who are actively evaluating AI tools, MCP servers, and dev infrastructure. Our weekly newsletter goes to engaged technical decision-makers.
All sponsored content is clearly labeled per our editorial policy.