OpenAI takes gold at math Olympiad, beating 90% of participants
By ForGeeks
OpenAI’s experimental artificial intelligence model achieved a gold medalist-level result at the 64th International Mathematical Olympiad (IMO), solving problems that require not only computation but also creativity.
Key details:
- 35 out of 42 points – that’s how many points the model scored after solving 5 out of 6 problems.
- Top 10% – this result surpasses 90% of human participants (in 2024, only 67 out of 630 mathematicians received gold medals).
- Unexpected breakthrough – AI usually does well in data processing, but loses in tasks where intuition and out-of-the-box solutions are needed.
“We created an algorithm that builds complex proofs like a professional mathematician,” wrote Alexander Wei, a researcher at OpenAI, in X.
What it means
- II learn to “think” creatively – not just looking for patterns, but finding new ways of doing things.
- But GPT-5 won’t have such capabilities – the company has no plans to introduce this technology into commercial products yet.
Context:
IMO is the most prestigious competition for young mathematicians. The tasks look short, but require deep reasoning – for example, to prove an inequality or find a hidden pattern.
What’s next?” OpenAI continues to develop AI’s “math thinking. The next goal is to solve all 6 problems and score the maximum points.
The story OpenAI takes “gold” at math Olympiad, beating out 90% of participants was first published on ITZine.ru.
