Humans outperform AI at this highly rigorous mathematics test

A new benchmark test designed to assess artificial intelligence systems on novel mathematics problems shows that humans still outperform AI. The benchmark, detailed in a study published online on June 12, 2026, in Nature, presented AI models and human mathematicians with previously unseen problems. The results indicate that while AI has made significant strides, it has not yet reached the level of expertise demonstrated by top human mathematicians when faced with entirely new mathematical challenges. This rigorous evaluation aims to provide a more accurate measure of AI's true mathematical reasoning capabilities beyond tasks it has been trained on. The findings suggest that current AI systems struggle with the kind of creative problem-solving and abstract reasoning that characterizes advanced human mathematical thought. Further research is needed to understand the specific limitations of AI in this domain and to develop methods for improving its performance on genuinely novel problems. The study's authors emphasize the importance of such benchmarks for tracking progress and identifying areas for future development in artificial intelligence research, particularly in fields requiring advanced cognitive abilities.