The unreleased reasoning model from OpenAI clinched a gold medal at the International Mathematical Olympiad (IMO), igniting significant drama within competitive math communities.

Amidst a typical academic break for most high school students, leading math students globally vied at the IMO, renowned as the top math competition. AI research labs also joined the fray with their LLMs, and an undisclosed OpenAI model achieved a score high enough to secure a gold medal, as researcher Alexander Wei revealed on X.

This OpenAI model addressed five out of six problems, amassing 35 points out of 42. “Three former IMO medalists evaluated the model’s submitted proof for each problem independently, with scores confirmed after unanimous agreement,” Wei noted. The problems posed algebra and pre-calculus puzzles that demanded innovative problem-solving, underscoring the model’s remarkable capacity to handle intricate proofs.

Nonetheless, the timing of this announcement received backlash for detracting from the achievements of human competitors. Reports indicate that the IMO had asked AI labs collaborating with them to postpone announcements by a week to prevent eclipsing the students’ accomplishments. This information surfaced from an X post by Mikhail Samin, who leads the AI Governance and Safety Institute nonprofit. OpenAI remarked that they did not engage formally with the IMO for result validation, opting instead to work with individual mathematicians and thus not obliged by any agreements. Mashable reached out to Samin via X for his comments.

Speculation suggests this unsettled organizers, who deemed OpenAI’s actions “rude” and “inappropriate.” These remarks are derived from hearsay shared by Samin, who also published a screenshot of a similar remark from an individual named Joseph Myers, presumably a two-time IMO gold medalist. Mashable contacted Myers for commentary, but he has yet to verify the authenticity of the screenshot.

In reply, OpenAI researcher Noam Brown asserted that they released results after the IMO closing ceremony, respecting an organizer’s wish. Brown further indicated that OpenAI was not in direct communication with IMO, suggesting no agreements concerning announcement timing.

Meanwhile, Google DeepMind allegedly collaborated with the IMO and declared that an “enhanced version of Gemini with Deep Think officially attained gold-medal standards at the International Mathematical Olympiad.” Per the announcement, DeepMind’s model was “officially assessed and validated by IMO coordinators using identical criteria as for student submissions.” The timing appears to be more than coincidental.

While some are captivated by the Real Housewives, the etiquette of elite mathematics competitions delivers the excitement we crave.

Disclosure: Ziff Davis, the parent company of Mashable, filed a lawsuit against OpenAI in April, claiming copyright infringement in training and operating its AI systems.