“Corporations will release a lightweight artificial intelligence (AI) model that can be utilized most universally without burdening graphics processing unit (GPU) resources as open source starting on the 24th.”
Kim Yu-won, CEO of Naver Cloud, noted during a press briefing held on the 23rd at Naver Square in Yeoksam-dong, Seoul, that as Naver's artificial intelligence (AI) strategy accelerates, they are producing competitive specialized models that can be applied stably and cost-effectively to Naver services, with the most in-demand models recently being lightweight models and inference models. He said, “This will contribute to activating the domestic AI ecosystem through the free sharing of the country’s first commercial large language model (LLM) as open source.”
Naver Cloud has decided to make three types of commercial open source available for free. Specifically, they are the lightweight models developed from the large language model (LLM) called ‘HyperCLOVA X’: HyperCLOVA X Seed 3B, HyperCLOVA X Seed 1.5B, and HyperCLOVA X Seed 0.5B. HyperCLOVA X Seed 3B is a visual model capable of processing image and video information. HyperCLOVA X Seed 1.5B is a model specialized in task execution capabilities. The most ultra-lightweight model, HyperCLOVA X Seed 0.5B, can implement a miniature conversational interface.
Kim emphasized the necessity of AI sovereignty in the context of the competition for technological supremacy between the U.S. and China. He noted, “As the competition intensifies and uncertainty grows, it can be difficult if we do not have the capability to create and independently operate AI,” and added, “As seen in the Ukraine-Russia war, ‘sovereign AI’ is not a choice directly related to security but a necessity.” He also pointed out, “Excluding the U.S. and China, South Korea is the only country with the overall capability to implement sovereign AI,” emphasizing the importance of a national all-out effort for this.
Kim criticized the movements demanding expansion of the certification to ‘high’ level, which deals with sensitive information such as defense and diplomacy, after global top three cloud companies, including Amazon Web Services (AWS), Microsoft (MS), and Google, obtained ‘level 1’ certification under the Cloud Service Security Certification Program (CSSP) to enter the domestic public market. He stated, “The private cloud market is fully open, and security is a minimal protection for national operations,” and emphasized, “It is not about the global companies being told to adjust to local regulations; rather, they need to adapt to local regulations themselves. Demanding changes to regulations is crossing the line.”
Regarding KT, which is promoting the development of ‘sovereign AI’ in collaboration with Microsoft (MS), Kim criticized by saying, “Bringing in foreign products and just putting our label on them is absurd.”
Kim explained about the possibility of implementing image generation features that have created a craze for the ‘Ghibli-style’ ChatGPT. He stated, “(Naver Cloud) has already secured image generation technology, but proactive investment is required to enter training,” and added, “The reason we have not possessed AI comparable to big tech is due to investment, not because of a lack of technical capability. The emergence of 5-6 models like DeepSeek in China is not simply because they have top-tier researchers.”
Regarding the plan to launch popularized services, Kim noted, “We are avoiding creating a $200 subscription model like ChatGPT,” adding, “Our main objective is to improve Naver services.”