According to a report by the Korean Times on October 25th, data shows that nearly half of the models newly released by local developers in Korea this year are based on Chinese foundational models. This has led to concerns that the Korean AI industry may be technically dependent on China.
According to a latest report released by the think tank Rockbridge Network Korea on the same day, from January to August this year, among the 1,845 models released by Korean local accounts on the Hugging Face platform, 49.3% were based on Chinese models, 35.9% were based on American models, and 14.9% were based on Korean local models.
According to the introduction, Alibaba’s Qwen is the most widely used foundational model, accounting for 46.8% of all models, which is 864 models in total. Google’s Gemma ranks second, accounting for 14.7%. Meta’s Llama ranks third, with a share of 13.8%. South Korea’s LG AI Research Institute’s EXAONE ranks fourth, with a share of 9.3%. Kakao’s Kanana accounts for 2.7%.
OpenAI's GPT, Anthropic's Claude and Google's Gemini are all closed-source models with their underlying weights not publicly released. In contrast, Alibaba's Tongyi Qianwen and Google's Gemma allow developers to download the models and modify them according to individual needs.
This type of model is called an “open-weight model”. Since the model weights are made public, Korean developers can more easily adapt it for specific purposes. For example, they can use Korean data for fine-tuning, without relying on cloud-based artificial intelligence services.
The report indicates that low costs and a wide variety of available models make Chinese models highly attractive to Korean developers.
The Chinese Ministry of Foreign Affairs previously stated that the development of artificial intelligence in China is a result of high-level technological self-reliance and self-strengthening, and it also benefits from China's consistent adherence to the principles of joint consultation, joint construction, and open cooperation. "We always believe that all countries should promote the open, inclusive, beneficial, and benevolent development of artificial intelligence, serving the well-being of all mankind."
