KHALI

SHARED STORY

Loading story



Get Khali



Open in app

Free on the App Store

#root { display: none !important; } html, body { height: auto !important; } body { overflow: auto !important; margin: 0; background: #07111f; color: #f4f4f5; font: 16px/1.6 system-ui, sans-serif; } article { box-sizing: border-box; max-width: 760px; margin: 0 auto; padding: 48px 24px 72px; } h1 { font-size: clamp(2rem, 7vw, 3.5rem); line-height: 1.05; } h2 { margin-top: 2rem; } a { color: #64b5ff; }

science · 3 outlets · 3 linked reports

AI Systems Win Gold-Medal Scores at the Math Olympiad for the First Time

2026-07-07

AI-assisted: summaries and questions may be incomplete or incorrect. Verify important details with the linked original reporting.

Summary

In 2026, both Google DeepMind and OpenAI unveiled AI models that scored 35 out of 42 on the International Mathematical Olympiad — a gold-medal-level result — the first time AI crossed that threshold.
DeepMind's advanced Gemini model wrote rigorous proofs directly from the problems, in natural language, within the 4.5-hour time limit, up from a silver-medal showing the year before.
Still, about 10% of human contestants earned gold and five scored a perfect 42, so the machines haven't caught the very best humans yet.

The question

AI matching Olympiad medalists shows machine reasoning is advancing faster than expected.

Original reporting

Open this story in Khali

Get Khali on the App Store