Recent unofficial benchmarks show frontier models like Claude Opus 5, GPT-5.6 variants, and China’s dots-note-3.0 achieving perfect 42/42 scores on the 2026 IMO problems held in Shanghai, building on 2025 results where OpenAI and DeepMind systems reached the 35-point gold threshold. However, these demonstrations remain non-official: they often involve extended compute, human-curated problem translation, or relaxed timing, falling short of IMO’s strict human rules requiring fully autonomous, timed, pen-and-paper-style solutions under identical conditions. Mathematicians continue to highlight gaps in reproducibility and research-level generalization, sustaining trader skepticism. With the contest concluded and no verified official gold entry, the 71% “No” probability reflects consensus that current capabilities, while impressive, have not yet met the market’s precise resolution criteria.