Rapid progress in large language model reasoning drove the 70.5% market-implied probability for an AI gold medal at the 2026 International Mathematical Olympiad. In July 2025, Google DeepMind’s advanced Gemini Deep Think and an OpenAI experimental model both scored 35 out of 42 on the official IMO problems—exactly the gold cutoff—by solving five of six questions through enhanced chain-of-thought and formal verification techniques. This marked a clear jump from the prior year’s silver-equivalent results, fueled by reinforcement learning pipelines and model scaling. With the 2026 contest scheduled for July in Shanghai, traders expect further iterative releases and new high-quality math datasets to push performance above the cutoff, though novel problem design or evaluation disputes remain possible swing factors.