Performance

Performance is about whether the editor stays usable when the document gets large. This page shows the latest 10k Plate vs Slate snapshot from the docs benchmark route and the command that reproduces it locally.

Latest Snapshot

This snapshot was captured on June 2, 2026 from the docs benchmark route in headless Chrome. The benchmark uses a 10,000-block document with chunked rendering enabled.

Workload	Slate	Plate core	Plate with basic plugins
Open a mixed document	524 ms	866 ms	881 ms
Type in a mixed document	26 ms	43 ms	35 ms
Open a code-heavy document	526 ms	837 ms	1101 ms

Read this as a workload snapshot, not as a universal editor score. Mount time, typing latency, plugin cost, and document shape are different performance questions.

Run The Benchmark

Start the docs app:

<CodeBlockCommand __bunCommand__="bun run --filter www dev" __npmCommand__="npm run dev --workspace www" __pnpmCommand__="pnpm --filter www dev" __yarnCommand__="yarn workspace www dev" />

Then run the public editor benchmark in another terminal:

<CodeBlockCommand __bunCommand__="bun run --filter www perf:editor:public" __npmCommand__="npm run perf:editor:public --workspace www" __pnpmCommand__="pnpm --filter www perf:editor:public" __yarnCommand__="yarn workspace www perf:editor:public" />

The runner writes a compact JSON summary to apps/www/.tmp/editor-perf-public-summary.json.

What It Measures

The benchmark keeps the public comparison small on purpose:

Measurement	What it tells you
Opening a mixed document	The cost of mounting a large editor with normal rich-text variety.
Typing in a mixed document	Whether the editor still responds under a large loaded document.
Opening a code-heavy document	The cost of a harder mark-heavy rendering profile.

Plate core isolates the editor wrapper cost. Plate with basic plugins adds the common block and mark plugins many editors start with.

Read The Numbers

Compare like with like:

• use mount numbers for startup cost
• use typing numbers for interaction cost
• use the basic plugin column when you care about a real Plate setup
• rerun the benchmark on your machine before publishing a performance claim

The raw output also includes lower-level Plate diagnostics, such as node-id cost. Those are useful for debugging Plate itself, but they are not the public story.

Source

Use the Huge Document demo for manual inspection. The benchmark route lives at /dev/editor-perf, and the runner lives in apps/www/scripts/run-editor-perf.mts.