Recent reports of frontier large language models such as GPT-5.6, Huawei’s Celia, and Xiaohongshu’s dots-note 3.0 reaching perfect or gold-medal scores on IMO 2026 problems reflect rapid gains in mathematical reasoning benchmarks. However, traders price “No” at 72 percent because of persistent gaps between unofficial benchmark results and verifiable, tool-free performance under official IMO grading rules. Key uncertainties include independent confirmation of claims, the stricter criteria applied to live contest problems versus training data, and the distinction between experimental systems and production models. Upcoming catalysts such as final official scoring releases and additional model disclosures could shift sentiment if they demonstrate sustained elite capability without human steering.