Resolution is based on the chatbot arena LLM leaderboard (https://lmarena.ai), specifically the company with the highest Arena Score in the Overall category, without style control or show deprecated, at the end of September 30th, 2025 11:59PM ET.
In the case of a tie, all companies tied for 1st place resolve to equal probability, such that they sum to 100%.
See also:
/Bayesian/which-company-has-the-best-ai-model
/Bayesian/who-will-have-the-best-texttoimage-SO0uN6suuS
/Bayesian/who-will-have-the-best-texttovideo-AtZ0CdIc8Z
/Bayesian/which-company-has-best-ai-computer
/Bayesian/which-company-has-best-vision-ai-en
/Bayesian/which-company-has-best-search-ai-mo
Previous months:
/Bayesian/which-company-has-best-ai-model-end
/Bayesian/which-company-has-best-ai-model-end-I0QsydsZuz
/Bayesian/which-company-has-best-ai-model-end-0CRdhqptRl
@JoshSnider neat! i'd be surprised if they released opus 4.2 before a new sonnet version (oh ig they did increase its context window to 1m, maybe that counts enough to make it not surprising)
