WebDev Arena

How it works

Arena.ai's WebDev Arena: two models each generate a web app from the same prompt, and a human picks the better result; votes aggregate into an Elo rating.

Every benchmark is scored on four weighted criteria — how directly it measures a real coding agent, how much of the stack it covers, how real its tasks are, and how open and reproducible it is. Those blend into the composite that ranks it #13 of 19.

Our review

Higher realism than chat-preference Elo because the models produce actual working web UIs that get judged — a real, if subjective, artifact. But it's single-shot preference voting (no iteration, no tools) on a proprietary board, so it's a useful signal for one slice of coding, weighted accordingly.