**Recent breakthroughs in AI mathematical reasoning have driven the 82% market-implied probability that an AI achieves IMO gold-medal performance in 2026.** In July 2025, Google DeepMind’s advanced Gemini with Deep Think and an experimental OpenAI model both reached the gold threshold of 35/42 points on the official IMO 2025 problems by solving five of six questions with rigorous natural-language proofs, matching or exceeding typical gold-medalist scores under contest-like conditions. This milestone followed DeepMind’s 2024 silver-level result and reflected rapid gains from improved large language models, specialized reasoning pipelines, verification-and-refinement techniques, and training on high-quality mathematical data. Subsequent releases—including GPT-5 and enhanced reasoning models—plus ongoing work on benchmarks like AIME have further strengthened capabilities. With IMO 2026 set for July in Shanghai and continued frontier-model progress expected before then, traders view sustained or improved performance as the base case, tempered only by uncertainties around exact resolution criteria, official verification, or edge-case problem difficulty. The 2025 results established the key precedent, shifting consensus sharply toward yes.