Remix 3's beta drops React for a Preact fork and a web-standards runtime

Remix 3 reached public beta on July 28, and the headline is not a new feature but a subtraction: the framework no longer runs on React. The beta preview is a ground-up rewrite built on a fork of Preact, an imperative state model, and Fetch-based routing, and it bundles far more of the stack than earlier Remix did. For teams that adopted Remix as a React framework, the practical question is now whether Remix is still a React framework at all, and the answer the project gives is no.

What Remix 3 actually is

Instead of React, Remix 3 renders through a forked Preact runtime and drops the declarative useState model for a direct one. State is a plain variable, and an explicit this.update() call signals that something changed and triggers a re-render. Routes are Fetch API routes whose controllers return standard web Response objects, so the server owns the request lifecycle in terms the platform already defines rather than framework-specific abstractions.

The most visible new primitive is what Remix calls “frames”: server-rendered fragments with a src attribute that load and reload independently, a server-driven UI approach comparable to how HTMX swaps in HTML from the server. On top of that, Remix 3 pulls a wide surface into the framework itself. Its own description lists routing, request handlers, middleware, sessions, auth, forms, uploads, assets, data and database management, UI components, theming, networking, and tests as core pieces rather than choices you assemble yourself. Per InfoQ, the preview sits at version 3.0.0-beta.5 with a weekly release cadence, and a new project starts from npx remix@next new.

The split that led here

This is the resolution of a reorganization that has been unfolding for over a year. When Remix and React Router converged, Remix v2’s capabilities were folded into React Router v7 under what the project calls framework mode, effectively making React Router the continuity path for the Remix philosophy in a React-native form. Remix 3 is the other branch: a fresh start that keeps the Remix name but leaves React behind. The two products are no longer versions of one lineage so much as two different answers to the same question, one that stays inside the React ecosystem and one that steps outside it.

That framing matters because it changes what “upgrading Remix” means. There is no incremental path from a Remix v2 app to Remix 3. Moving to Remix 3 is a rewrite onto a different runtime and a different component model, not a version bump.

What it means for teams today

For the large base of existing Remix and React shops, the near-term guidance is unglamorous: stay where you are. React Router v7 carries the proven Remix patterns forward in React and remains the stable option, and Remix 3 is explicitly early. LogRocket’s assessment is blunt, calling it “research, not production-ready” and advising teams not to bet a product on it yet. A beta preview on a weekly release cadence, with a component model that most React developers have never written, is a thing to prototype with, not to migrate a shipping app onto.

The more interesting decision is for teams starting fresh and weighing how much JavaScript they want to ship. Remix 3’s Preact base and server-driven frames point at a lighter client runtime than a full React meta-framework, which is a real consideration for content-heavy or performance-sensitive sites. That is a prototype-and-measure question today, not a commitment, but it is the reason the release is worth tracking rather than dismissing as a rename.

The wider signal

A framework with Remix’s visibility choosing to leave React is a notable data point in a broader drift back toward web-platform primitives and less client-side JavaScript, the same current that runs through HTMX, Astro’s islands, and the general renewed interest in server-driven UI. It does not signal that React is in retreat; React still anchors the ecosystem, and React Router inherits most of Remix’s existing users. What it does suggest is that the meta-framework layer is willing to reconsider its foundation when the platform offers enough on its own. The signal to watch is adoption: whether Remix 3 attracts real production usage as it stabilizes, or whether the audience that valued Remix mostly follows the React Router path. Stackmaven will check back on or around October 28.

Sources cited

1. Remix 3 Beta Preview (Remix Blog) remix.run
2. Remix 3 Beta Preview Ditches React for a Web-Standards Full-Stack Framework (InfoQ) www.infoq.com
3. Remix 3 ditched React: Should you stick with it? (LogRocket) blog.logrocket.com