The rapid proliferation of AI chatbots has made it difficult to know which models are actually improving and which are falling behind. Traditional academic benchmarks only tell you so much, which has ...