Which company has the best AI model end of September?

What this question means

Anthropic at 87%, OpenAI at 10%, Google at 5%, with Meta and SpaceXAI at 1% and the remainder of a twenty-one name field at zero.

Set that against the Code Arena WebDev question elsewhere on this board, where OpenAI leads at 68% and Anthropic sits at 28%. Same companies, same month, opposite ordering — because the two questions are scored on different things. One measures a specific coding evaluation; this one is broader.

There is a general point here about benchmark questions. The apparent disagreement between these two markets is not the crowd being inconsistent. It is the crowd correctly pricing two genuinely different tests. A model can lead comfortably on one measure and trail on another, and any question resolving on "best" is really a question about which yardstick is being used.

Read the resolution criteria before the leaderboard. It decides more than the models do.

Settles at the end of September.