Which company has the best Code Arena WebDev AI model end of September?

What this question means

OpenAI at 68% leads this one, with Anthropic at 28% and everyone else — Alibaba at 1%, then Moonshot, Z.ai, SpaceXAI, Meta and ByteDance — at or near zero across a sixteen-name field.

The comparison worth making is with the broader question on this board asking which company has the best AI model overall at the end of September. There the ordering reverses sharply. Same companies, same deadline, different leader.

That is not a contradiction — it is the whole point of a benchmark. This question is scored on Code Arena's WebDev evaluation specifically, a narrow measure of one capability. The general question aggregates differently. Models are not uniformly better or worse than one another; they are better at particular things, and which benchmark you pick decides the answer.

Anyone forecasting either should read the resolution source first. On questions like this the scoring method carries more weight than any judgement about the models themselves.