Evaluating current online AI models through human preference is the promise, and it does it well. It takes all most recent AI online tools and rank them according to several criteria and purposes, such as: overall performances, math, creative writing, coding and much more, including the language used.
Yes, this is human-vote driven, and thus subject to some external influences, but the authority of the site and the number of voters make that evaluation very relevant and up-to-date.
