🏅 Top traders

Extra decimal view

wait what just happened

@AndrewMackenzie69e9 guess just adjustment by additional scores? Well within error bars, but boy I really did not expect this to happen.

@HenriThunberg no it's a slightly new model. the first score was a preview-preview. it seems they are continuing to post train it

This resolves based on the scores in one week.

case for yes:

grok 3 is only 26 points above chatgpt-4o-latest

OpenAI testers preferred 4.5 at a 57-63% rate vs GPT-4o depending on category

Those testers might be different from Chatbot arena but if they're similar, 4.5 will crush.

@JoshYou I should trade with more conviction lol

OpenAI announces GPT-4.5, warns it’s not a frontier AI model

system card seems to imply a lot less than 85% chance of this. MMLU score is 88%, vs grok 3 93%. and grok has clearly done something to try to optimize for lmarena, which i don't think openai will care about

@jim if one of them doesn't show up on there, does this resolve NO? or stays open indefinitely