AI Times, 10 Jun 2025
The Korean tech landscape has witnessed a surge in AI development, with established players like Naver and Kakao aggressively pursuing large language models (LLMs). Now, a new contender emerges: Xiaohongshu, the popular social media platform known as “Red” or “Little Red Book” internationally. According to the article, Xiaohongshu has unveiled its own LLM, dubbed “dots.llm1,” through open-source platforms like Hugging Face and GitHub.
This move is particularly interesting given the current global trend towards open-sourcing LLMs, spearheaded by companies like Meta and Google. In Korea, Naver has also been actively contributing to the open-source community with projects like HyperCLOVA. Xiaohongshu’s entry signals a growing recognition of the benefits of collaborative development and wider accessibility in the AI field.
The article notes that dots.llm1 employs a “Mixture of Experts” (MoE) architecture, activating only 14 billion out of its 142 billion parameters. This approach, increasingly popular in the LLM space, optimizes computational efficiency without significantly compromising performance. Similar strategies are being explored by other Korean tech companies like Kakao Brain, highlighting the industry’s focus on balancing performance with resource utilization.
From a technical perspective, the MoE architecture allows the model to dynamically select the most relevant “expert” sub-networks for a given task. This specialization enhances efficiency and allows for scaling to massive parameter counts, which is crucial for complex language understanding. Given the nuances of the Korean language and the rapid evolution of online slang, this approach could prove particularly beneficial for applications tailored to the Korean market.
The Korean government’s increasing focus on AI regulation and ethical considerations will undoubtedly impact Xiaohongshu’s LLM deployment strategy within the country. Furthermore, the competitive landscape, with major players like Naver and Kakao vying for dominance, will present a significant challenge. It remains to be seen how Xiaohongshu will differentiate its offering and navigate the regulatory complexities of the Korean market.
Xiaohongshu’s foray into the open-source LLM arena raises several questions. Will this move stimulate further innovation within the Korean AI ecosystem? How will it influence the ongoing debate around open vs. closed AI models? And, perhaps most importantly, how will Xiaohongshu leverage its LLM to enhance its platform and user experience in the face of intensifying competition?