Current LLMs struggle to solve formal mathematical conjectures that lack existing proofs. While models excel at pattern recognition, they fail at the novel logical leaps required for Millennium Prize Problems. This gap confirms that scaling data alone won't replace human mathematical intuition. Researchers must find new ways to implement rigorous reasoning.