Chinese technology company Xiaohongshu, internationally known as RedNote, has announced a historic breakthrough in artificial intelligence reasoning. The company’s dots-note 3.0 model achieved a perfect score of 42 points on this year’s International Mathematical Olympiad (IMO) problems, making it the first AI system in the competition’s history to secure both a full score and a gold medal-level result.
The achievement places dots-note 3.0 in an elite category, as it is only the second AI model to receive an officially graded gold-level result, following Google DeepMind’s Gemini Deep Think, which scored 35 points in 2025. The IMO evaluation is particularly rigorous because it requires complete, logical proofs rather than just final answers. A solution can lose points if any step in the reasoning process is unsupported, making the contest a definitive test of an AI’s abstract reasoning capabilities and the consistency of its arguments.
The success of dots-note 3.0 represents a significant leap forward in the quest for artificial general intelligence (AGI). Mathematical reasoning has long been considered a critical benchmark for AI systems, as it requires the ability to understand abstract concepts, formulate logical arguments, and execute multi-step problem-solving strategies. By achieving a perfect score on the IMO, Xiaohongshu has demonstrated that its model possesses reasoning capabilities that rival those of the world’s most talented human mathematicians.
Agent-Based Reasoning System
Xiaohongshu revealed that dots-note 3.0 utilized an advanced agent-based reasoning system to generate its mathematical proofs. The six problems evaluated covered complex topics including algebra, geometry, number theory, and combinatorics, with each problem worth seven points. During the evaluation process, human intervention was strictly prohibited; researchers were not allowed to adjust the model’s reasoning, provide hints, correct answers, or select among multiple outputs.
The use of an agent-based system is particularly noteworthy. Unlike traditional large language models that generate responses sequentially based on statistical probabilities, agent-based systems are designed to break down complex problems into smaller, manageable tasks. They can explore multiple solution paths, evaluate the validity of intermediate steps, and backtrack if they encounter a logical dead end. This approach is essential for solving the type of open-ended, multi-step problems found in the IMO.
The version of the model used for the IMO evaluation currently remains internal to the company. However, Xiaohongshu has indicated that it expects to release the model as open-source to the broader AI research community in the near future, which could significantly accelerate global research into AI mathematical reasoning.
A Rare Achievement
The perfect score achieved by dots-note 3.0 is exceptionally rare, even among the world’s top human mathematical talent. Of the more than 600 students who participated in the official IMO competition, only seven individuals, approximately one percent, received a perfect 42 points.
The gold medal cutoff for this year’s competition was set at 29 points, meaning the AI model’s performance exceeded the gold threshold by a substantial 13-point margin. This milestone demonstrates the rapid progress Chinese AI laboratories are making in developing models capable of complex, multi-step logical reasoning, challenging the dominance of Western tech giants in the pursuit of artificial general intelligence.
The achievement is particularly impressive given that Xiaohongshu is primarily known as a social media and e-commerce platform, rather than a dedicated AI research lab like DeepMind or OpenAI. It is also consistent with the broader trend of many Chinese AI companies pivoting from the frontier model race to harder, more specialized capabilities.
Implications for AI Development
The success of dots-note 3.0 has profound implications for the future of artificial intelligence development. As AI models become increasingly capable of advanced mathematical reasoning, they will be able to tackle a wider range of complex problems in fields such as physics, engineering, computer science, and finance. This could lead to significant breakthroughs in scientific research, technological innovation, and economic productivity.
Furthermore, the open-source release of dots-note 3.0 could democratize access to advanced reasoning capabilities, enabling researchers and developers worldwide to build on Xiaohongshu’s work. This collaborative approach could accelerate AI innovation and foster the development of new applications and services that leverage advanced mathematical reasoning.