Plate vs Lexical: components included or nothing included

What each one decides for you

Plate decides a great deal: which plugins exist, what a toolbar looks like, how block menus behave. You can change all of it, because the components live in your repository, but you begin with opinions.

Lexical decides almost nothing beyond the state model. Every visible element is yours, and the framework's position is that this is correct.

Our benchmark measures the consequence: thirteen lines against twenty-five, for editors that render the same thing.

Where Plate wins

A running start. Components, plugins and patterns for the features document editors need — mentions, tables, slash commands, drag handles.

Design-system fit. If you use shadcn/ui, the editor matches your product immediately.

AI plugins in the maintained set, rather than as a service or a project.

A larger catalogue of ready node types, which shortens the distance to a product rather than a prototype.

Where Lexical wins

Nearly a third less weight, at 104.9 kB against 150.8 kB before either adds plugins.

A cleaner state model. Immutable and transactional, where Slate's normalisation is subtler and harder to predict at the edges.

Meta's production scale behind the core.

Fewer assumptions to unwind when your editing surface is not a conventional document.

No design-system coupling. Plate's advantage becomes a constraint if you are not on Tailwind and shadcn/ui.

The upgrade question

Both projects move. Plate publishes major versions frequently; Lexical is pre-1.0 and permits API changes by convention.

The practical difference is what breaks. Plate's churn tends to touch components you copied, which is your code and therefore your merge. Lexical's touches framework APIs you called, which is a smaller surface but a harder fix.

Whichever you choose, treat editor upgrades as scheduled work rather than as an occasional inconvenience. Teams that do not end up pinned to an old version with a growing list of unbackported fixes.

Migration between them

Both build structured documents, so content converts with a serialiser you write. Neither's plugin API resembles the other's, so the editor layer is a rewrite.

Realistically this is a decision to make once. Pick on the basis of how much editor you intend to own, not on which API reads better in a tutorial.

Decision checklist

1. Are we on Tailwind and shadcn/ui? That alone justifies Plate for many teams.
2. How unusual is our editing surface? Unusual favours Lexical.
3. What is our JavaScript budget? The gap is 45.9 kB before plugins.
4. Who owns editor upgrades, and when?
5. Do we need collaboration? Both can; neither hands it to you free.

The number that matters

45.9 kB — the gzip gap in our bundle benchmark, and roughly the weight of the components Plate gives you. That is a fair way to read this comparison: you are paying in kilobytes for the UI you would otherwise build, and the question is whether you would have built the same thing.