weather-card-benchmark

Identify your card

Built your own card from the benchmark prompt? Paste its HTML and see whose code yours looks like — matched against every model in the set on code-structure alone.

What this measures — and what it doesn't

This compares only the code structure of your card — its markup, CSS, and JS — never how it looks. Nothing is rendered and no pixels are judged, so read the result as “whose code does mine resemble,” not “which model made it.” It also runs a structural subset of the benchmark's channels, not the full pipeline.

Your HTML never leaves your browser — the whole comparison runs client-side against a ~0.5 MB signature file.

Which prompt did you build from?