Humanity’s Last Exam is a large-scale collaborative benchmark designed to test the limits of frontier AI systems on the hardest questions across diverse domains of human knowledge.