Xiaohongshu AI team open-sources dots3-note, scores 75.1 on Terminal-Bench 2.1
- Xiaohongshu AI Lab's dots.studio announced the open-sourcing of dots3-note preview on August 15.
- The model uses a 280 billion parameter MoE architecture with 16 billion active parameters.
- It supports a 512,000-token context window and multimodal understanding across text, vision, and audio.
- The model introduces the TEMPO reinforcement learning method for long-horizon agent training.
- A SemiAnalysis chart showed dots3-note scoring 75.1 on Terminal-Bench 2.1, 4.9 points above the best-performing U.S. open-weight model in the chart.
Xiaohongshu AI Lab's dots.studio said it has open-sourced dots3-note preview. The model is described as a 280 billion parameter MoE system with 16 billion active parameters, a 512,000-token context window, and multimodal capabilities covering text, vision, and audio.
The team said the model introduces a TEMPO reinforcement learning method for long-horizon agent training. According to the report, the model weights are already available on Hugging Face and its API has been integrated into OpenRouter.
A chart shared by SemiAnalysis showed dots3-note scoring 75.1 on Terminal-Bench 2.1, 4.9 points higher than the best-performing U.S. open-weight model in that chart. SemiAnalysis said it is still testing the model in everyday use to judge practical quality and whether benchmark over-optimization is present.
For markets, the item is likely neutral overall, with limited direct impact on stocks, crypto, gold, or foreign exchange beyond the usual attention on AI model development and open-weight competition.
小红书AI团队开源 dots3-note，Terminal-Bench 2.1 得分75.1
小红书 AI Lab 的 dots.studio 表示，已开源 dots3-note preview。报道显示，该模型采用 2800 亿参数的 MoE 架构，其中 160 亿参数处于激活状态，支持 51.2 万 token 上下文窗口，并具备覆盖文本、视觉和音频的多模态理解能力。
团队称，该模型引入了 TEMPO 强化学习方法，用于长时程智能体训练。报道还提到，模型权重已在 Hugging Face 上提供，API 也已接入 OpenRouter。
SemiAnalysis 发布的图表显示，dots3-note 在 Terminal-Bench 2.1 上得分 75.1，比图表中表现最好的美国开源权重模型高 4.9 分。SemiAnalysis 也表示，团队仍在日常使用中测试该模型，以评估实际质量以及是否存在针对基准的过度优化。
从市场角度看，这则消息整体偏中性，对股票、加密货币、黄金或外汇的直接影响有限，主要反映的是 AI 开源模型竞争的进展。