Recent official and independent evaluations of 2026 IMO problems show frontier large language models, including Claude Opus 5, GPT-5.6 variants, and Chinese systems like Huawei Celia and Xiaohongshu dots-note-3.0, achieving perfect 42/42 or gold-threshold scores of 30+/42 under timed, natural-language conditions. Open-weight pipelines such as NVIDIA Nemotron 3 Ultra and DeepSeek V4 Flash have also cleared the gold cutoff with reproducible inference methods. Despite these benchmarks, trader consensus heavily favors No at 86% implied probability because no AI has secured an official gold medal through formal IMO entry, judging, or the designated Grand Challenge resolution process. Key catalysts include ongoing debates over grading verification, lack of direct competition against human contestants, and strict resolution criteria that distinguish benchmark saturation from contest victory. Upcoming reports on official recognition could still shift sentiment before year-end.