Benchmarks

Each framework renders the same app: a toolbar and a table of 1,000 keyed rows with identical, seeded labels. The implementations are written the way each framework documents — keyed lists, memoized rows where the framework offers them — with no DOM work outside the framework.

A headless Chrome runner clicks the same buttons a user would. Each operation is timed in the page from just before the click until a forced synchronous layout after the microtask queue drains, so the number covers script, style and layout and excludes paint. Every operation runs 5 untimed warmups and 15 timed samples per framework, in round-robin order across the frameworks on a fresh page per operation; the table shows the median.

Startup is the time from navigation start to the frame after the app mounts, over 10 cold loads. Memory is the JavaScript heap after rendering 1,000 rows and a forced garbage collection. Bundle size is the minified JavaScript each app ships, raw and gzipped; the shared stylesheet is excluded.

The numbers are remeasured when a framework version or loom itself changes, on the machine named above. Absolute times move with the hardware; the ratios to vanilla JavaScript are the comparable part.

The bench app, runner and results schema are in the repository: apps/bench.