Comment by surround

You may have already figured this out, but the leaderboard you linked to (https://lmarena.ai/leaderboard/text/overall-no-style-control) shows gemini-2.5-pro ahead with a score of 1471 compared to gpt-5 at 1462.