This market will resolve to the year when a freely available large language model achieves a gold medal-level performance at the International Mathematical Olympiad. The model must be available before day 1 of the IMO. I must be able to use the model without paying.

For each year, directly after each day of the exam, I will pick the strongest LLMs and give them the problems independently. Each LLM has one attempt at each problem. I will then grade the problems. I will not be using the IMO rubrics. Grading will be based on these rough guidelines for USA olympiads: 7 for a complete solution, 6 for an easily fixable tiny, 5 for a small non-central mistake, 1 for nontrivial progress, and 0 for no nontrivial progress towards solving the problem. In particular, most humans would receive 7 for a correct solution and rarely receive more than 2 points for an incomplete solution.

The market will resolve to the first year which any model scores at or above the gold medal cutoff. If this does not occur at or before the 2029 IMO, the market will resolve to 2030 or later.