Recent breakthroughs in AI mathematical reasoning have driven the 73.5% implied probability for an AI system claiming gold at the 2026 International Mathematical Olympiad. Google DeepMind’s Gemini Deep Think and OpenAI’s reasoning models both scored 35 out of 42 points on the 2025 IMO problems—precisely the gold threshold—by solving five of six questions under human-like conditions. This marked a sharp advance from DeepMind’s 28-point silver-level result on the 2024 set, reflecting gains in formal verification, reinforcement learning, and large language model scaling. With the 2026 contest scheduled for July in Shanghai and fresh training datasets such as MIT’s MathNet now available, traders see continued iteration as likely to clear the cutoff, though harder problem sets or evaluation disputes could still shift outcomes.