Hacker News Reader

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.

It is very interesting that among top-20 models, all non-proprietary ones are from China.

gpt-5 was ahead on that last night

The leaderboard hasn't changed since it was updated to add gpt-5. Here's what it looked like yesterday https://archive.is/XIrbN

If you saw gpt-5 was ahead, you might have been looking at the leaderboard with style control https://lmarena.ai/leaderboard/text/overall