OpenAI's new language model "o1" has surpassed human performance in complex reasoning tasks, such as math, coding, and science.