Overview
dots is a large model series from the Xiaohongshu technical team. Its internal version dots-note 3.0 achieved a perfect score on all six problems at the 67th IMO 2026, earning a gold medal with 42/42 points, a feat accomplished by only 7 human contestants worldwide. The model does not rely on formal language; it directly reads original LaTeX problems and completes problem-solving end-to-end through recursive self-critique capabilities. dots-note 3.0 is the lightest model in the dots3 series and is expected to be open-sourced.
Key Features
- IMO Perfect Gold Medal: dots-note 3.0 scored a perfect 42/42 on all six problems at IMO 2026, a result achieved by only 7 human contestants globally
- End-to-End Problem Solving: Does not rely on formal language; directly reads original LaTeX problems
- Recursive Self-Critique: Completes multi-step reasoning through recursive self-critique capabilities
- Lightweight: dots-note 3.0 is the lightest model in the dots3 series
Use Cases
- Mathematical competitions and high-difficulty reasoning
- Scientific research reasoning tasks
- Formal reasoning research
Pros
- IMO perfect score validates reasoning ability
- End-to-end without relying on formal language
- Lightweight and expected to be open-sourced
Pricing
dots-note 3.0 is expected to be open-sourced; specific details are subject to the announcement by the Xiaohongshu technical team.
Summary
dots is a large model series from Xiaohongshu. dots-note 3.0 demonstrates end-to-end mathematical reasoning ability with a perfect gold medal at IMO 2026. Its distinctive feature is not relying on formal language, making it suitable for high-difficulty reasoning research scenarios.
Version History
- 小红书开源连续自回归语音合成模型 dots.tts：打造可持续扩展的 TTS 基座 (2026-08-13): Xiaohongshu dots team open-sourced the 2-billion-parameter fully continuous end-to-end autoregressive speech synthesis model dots.tts, achieving the best average content accuracy and average speaker similarity across three subsets of Seed-TTS-Eval.
- dots3-note Preview 开源：280B 参数轻量模型，主打长程智能体与多模态推理 (2026-08-14): Xiaohongshu Technology open-sourced dots3-note Preview, the lightest model in the dots3 series, with a total of 280B parameters and 16B activated parameters, supporting 512K context and multimodal understanding of text, vision, and speech, optimized for complex reasoning and long-horizon Agent tasks.
- 小红书 dots 模型获 IMO 2026 满分金牌 (2026-07-21): Xiaohongshu's Dots team participated in the 67th IMO 2026 with their internal version dots-note 3.0, achieving full marks on all six problems with a score of 42/42, earning a perfect gold medal. Only seven human contestants worldwide achieved this result. The model does not rely on formal language; it directly reads original LaTeX problems and completes problem-solving end-to-end through recursive self-critique capabilities. dots-note 3.0 is the lightest model in the dots3 series and is expected to be open-sourced.
- dots-note 3.0 IMO Perfect Score (2026-07-21): dots-note 3.0 achieved a perfect 42/42 gold medal at IMO 2026, solving problems end-to-end without formal language, and is expected to be open-sourced