]
) OpenAI achieved
at the International Math Olympiad. This is another important milestone in AI’s rapid growth. Alexander Wei is a research scientist working at OpenAI on LLMs, reasoning and other topics. Posted on X an experimental research model delivered this “longstanding grand challenge in AI.”
Wei said that an unreleased OpenAI model was able solve five out six problems in one of the oldest and most prestigious math competitions in the world, earning 35 points out of 42 total. In the International Math Olympiad, countries send up six students to solve very difficult algebra and precalculus problems. The exercises may seem simple, but they require creativity to get the best marks. Only 67 out of 630 contestants won gold medals in this year’s competition. That is about 10%.
AI’s are often tasked with dealing with complex datasets and repetitive tasks, but they fall short when it comes solving problems that require creativity or complex decision making. OpenAI claims that its model was able, in the latest IMO competition to solve complex math problems using human-like reasoning. Wei wrote
“By doing so, we’ve obtained a model that can craft intricate, watertight arguments at the level of human mathematicians,” on X. Wei, and Sam Altman CEO of OpenAI both added that they don’t expect the company to release anything with such a high level of math ability for several months. The GPT-5 is likely to be an improvement over its predecessor, but will not feature the same impressive capabilities that it needs to compete in IMO.
