Benchmarks

Measures gzip-compressed client JavaScript shipped for a route with no user-authored interactivity.

Measures gzip-compressed client JavaScript shipped for a minimal button-and-state interactive route.

Measures the same interactive route while opting out of optional client navigation runtime where the framework supports it.

Renders a production app route that emits 1,000 simple text spans.

Streams a production app route with 1,000 simple text spans and validates the complete response body.

Measures elapsed time until fetch resolves response headers for the real streaming route.

Measures elapsed time until the first response body chunk arrives for the real streaming route.

Measures elapsed time until the complete real streaming response body is consumed and validated.

Measures complete response latency for a route whose body waits on a 50 ms async boundary.

Measures complete response latency for two sibling 50 ms async boundaries; parallel renderers stay near one boundary.

Renders a static-cacheable app route with 1,000 simple text spans after the production server has warmed it.

Renders 200 cells with many dynamic escaped attributes, inline style values, and text content.

Renders a dynamic route that combines route parameters with server data before producing HTML.

Runs a fixed burst against the production fixture with up to 100 concurrent requests and reports sustained request throughput.

Runs the same concurrent request burst and reports per-request p99 latency, exposing event-loop stalls hidden by sequential tinybench runs.

Reports RSS growth across the concurrent request burst so sustained-load memory trends are visible in router benchmark output.

Measures browser client navigation to a route with loader data, covering data-bearing SPA transitions.

Measures browser back-forward restoration after SPA navigation so history snapshot regressions are visible.

Measures route-to-route client navigation latency in a real browser when the adapter provides a browser probe.

Measures page load time until the interactive route is visible and idle before any user interaction.

Measures the first click-to-visible-update latency immediately after DOMContentLoaded without waiting for network idle.

Measures the first click-to-visible-update latency after the interactive route has reached network idle.

Measures the second click-to-visible-update latency after the route has already handled one client interaction.

Measures gzip-compressed HTML payload bytes for the 1,000-node SSR route, complementing client bundle size cases.

Measures gzip-compressed production build output size when the adapter exposes build artifacts.

Loads an app route with 100 independently interactive islands and reports time until all islands can update in real Chromium. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Starts the framework dev server for a minimal app and reports server readiness latency. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Requests a minimal app route from a warm dev server and reports first request latency. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Edits a route module while the dev server is running and reports time until the changed response is observable. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Builds a 1,000-route app and reports request latency for a route near the end of the route table. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Builds a 1,000-route app and reports production server cold-start latency for that route scale. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Reports production build time for a 1,000-route app to catch route-count scaling regressions. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Reports process RSS growth while building and serving a 1,000-route app. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Renders a form with an inferred server action, submits the encoded form POST, and reports action roundtrip latency. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Renders a route under five nested layouts, guarding against sequential layout shell regressions. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Builds the Cloudflare Pages worker bundle and reports request latency through its exported fetch handler. A workerd/Miniflare harness should replace this fallback once the local workerd path-resolution failure is fixed. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Measures production server cold-start latency when the adapter can isolate startup from build work. This section currently compares mreact app-router variants only; it is not a cross-framework ranking.

Reverses 1,000 keyed rows and verifies that DOM node identity is preserved.

Creates 1,000 button event targets and measures initial interactive wiring cost without dispatching events.

Reports heap growth after repeatedly creating, updating, and clearing 1,000-row lists.

Updates 1,000 fine-grained source values when each source has one live non-DOM subscriber, separating direct source write overhead from aggregate computed fan-in and framework-level array update work.

Updates one reactive text value that is bound to 1,000 text nodes.

Updates one source value that fans out through a derived value into 1,000 displayed text nodes.

Updates the inputs feeding one aggregate and validates one derived aggregate text output. Caveat: this is not a direct cross-framework source-write comparison because mreact, Solid, and Solid v2 update 1,000 fine-grained sources, while React, Marko, and Qwik update one array/props payload.

Updates 1,000 fine-grained source values without subscribers, derived values, DOM writes, or framework-level re-render work, then validates the final source values. Frameworks without an equivalent source primitive report this case as unsupported.