Recent breakthroughs at the 2025 International Mathematical Olympiad, where Google DeepMind's Gemini Deep Think and an OpenAI experimental reasoning model each scored exactly 35 out of 42 points by solving five of six problems using natural-language proofs, have driven the 73% implied probability for an AI gold medal in 2026. This marked a sharp leap from the prior year's silver-level performance and reflected gains in large language model scaling, reinforcement learning for formal verification, and general-purpose reasoning pipelines. With iterative releases expected before the July 2026 contest in Shanghai, traders anticipate further capability gains across frontier labs, though factors like novel problem difficulty or evaluation disputes could still influence the outcome.