World's First AI Systems to Score a Perfect 42/42 at IMO 2026 Are Chinese

My take: For the first time in the history of the International Mathematical Olympiad, AI models reached a perfect score: 42 out of 42 points. Two models were officially graded by IMO organizers: Huawei's Celia and Xiaohongshu's dots-note 3.0 (known in the West as RedNote). Four other systems, including Claude Fable 5 and GPT-5.6 Sol, also reported 42/42, but in a test organized by a venture capital investor and graded by AI agents, not by official IMO judges. That distinction matters if you want to take the number seriously.

The fact that AI can solve Olympic-level math problems is a genuine advance in formal reasoning. The fact that the first to do so with official certification are two Chinese companies, rather than OpenAI, Google, or Anthropic, says something about the pace at which the global AI ecosystem is expanding beyond Silicon Valley.

For anyone who uses AI in analytical, technical, or mathematical work: the reasoning capabilities of frontier models are real and already exceed most human experts on very specific formal tasks. What processes in your work or business depend on advanced mathematical or logical reasoning that AI could already execute more efficiently?