Artificial Analysis's new blind-preference leaderboard for voice agents found the model people rate as most likeable isn't the one that best completes tasks.
Continue to AI University →