MANY MINDEDRedNote's beta AI model reaches a flawless Math Olympiad result
RedNote, the Chinese company behind the social platform Xiaohongshu, says its dots-note-3.0 model scored a perfect 42 points at the 2026 International Mathematical Olympiad, solving all six problems. According to the report, this is the first time an AI has achieved a perfect IMO score. This year's competition concluded in Shanghai.
The result matters because of what the IMO demands: not answers alone, but rigorous written proofs. The contest has run annually since 1959 and rewards multi-step reasoning rather than pattern matching. For comparison, at the 2025 IMO both Google DeepMind and OpenAI reached gold-medal-level performances of 35 out of 42. dots-note-3.0 is described as the lightest model in RedNote's dots3 family, which also includes larger models named jazz and aria tuned for different compute costs.
Rigorous machine reasoning is a general-purpose capability. If a small, efficient model can carry proof-level mathematics, the same reasoning could eventually reach tutoring, verification, and research assistance at low cost. RedNote also said it plans to open-source dots-note-3.0, which would put the capability in more hands rather than behind a paywall.
The caveats are real: the model is still in beta, and RedNote gave no firm timing for the open-source release. Independent verification of the score, and a look at the model once weights are public, will show whether the result holds up outside the company's own account.
Source: SCMP Tech