Huawei and Xiaohongshu’s AI systems achieved a perfect score at the International Mathematical Olympiad, marking the first time technology has matched top-tier human performance in the competition. This milestone indicates a significant leap in machine reasoning and logic capabilities, as multiple global labs now demonstrate similar problem-solving proficiency.
Detailed Coverage
Artificial intelligence has reached a major technical milestone by achieving a perfect score at the International Mathematical Olympiad, the world's most challenging competition for young mathematicians. For the first time, AI systems developed by Chinese firms Huawei and Xiaohongshu successfully solved all problems in the contest, matching the performance of the highest-ranked human participants.
Advancing Logic and Reasoning
The competition, held in Shanghai, tested participants on complex mathematical problems that require high-level logical reasoning rather than simple pattern recognition. Huawei’s Celia and Xiaohongshu’s dots-note-3.0 were subjected to the same official grading process as the 666 human contestants. Under strict rules, the AI systems were provided with the exam questions only after human participants had finished their tests, and no human assistance was allowed during the AI's processing phase.
This success highlights a rapid evolution in AI capabilities. While AI systems have previously earned gold-medal scores, reaching a perfect 100% marks a shift in how machines approach multi-step reasoning. Xiaohongshu noted that their dots-note-3.0 model, participating for the first time, represents a significant improvement over earlier large language models that struggled to maintain consistent logic across all exam questions.
Global Competition and AI Development
The ability to solve these advanced mathematical problems is now a key benchmark for technology companies globally. The progress is not limited to Chinese firms; other major players have also reached high performance levels. Recent data indicates that several cutting-edge AI models from US-based labs OpenAI and Anthropic, alongside domestic startup Axiom Math and China's Moonshot AI, also attained perfect scores of 42 out of 42 points on this year's exam questions.
For context, this represents a major leap from 2024, when Google's AI model achieved a silver-medal score by correctly solving four out of the six problems. As these models move beyond basic language tasks into complex logical fields, the focus for developers is shifting toward enhancing reliability in reasoning and reducing error rates in highly technical domains. Investors and industry analysts are now looking at how these enhanced problem-solving capabilities can be integrated into commercial applications, such as automated scientific research, complex engineering design, and advanced data analytics, which could eventually impact productivity and software development costs across the tech sector.