Neel Somani, a former researcher in the field of quantification, tested the mathematical abilities of the OpenAI GPT-5.2 model and was surprised to find a complete solution for a difficult mathematical problem.

The model's reasoning process was sophisticated, using advanced concepts and referring to previous discussions in the mathematical community. Additionally, the solution provided by AI was different and more comprehensive than previous approaches. These advances suggest that AI is becoming a valuable tool in mathematics, with 15 problems on the Erdős site moving from "open" to "solved" in just the last few months, many of the solutions crediting AI models. Leading mathematicians, such as Terence Tao, are closely observing these developments, acknowledging the significant contributions of AI in solving lesser-known problems. Moreover, an increased emphasis on formalization has helped facilitate these advances, and modern proof-assistance tools have lowered the barriers to entry. Although AI does not autonomously do mathematics, its role in solving long-standing problems suggests that the frontiers of mathematical discovery are changing rapidly.