The artificial intelligence (AI) solution market in Japan is expected to grow to 17 trillion won by 2030. The market for electronic document processing using AI, which is the main business of Upstage, is also about 10 times larger than that in South Korea.
Kim Sung-hoon, the CEO of Upstage, held a press conference at the Conrad Hotel in Yeouido, Seoul, on the 16th and explained the reasons for establishing Upstage Japan in March this year. Kim said, "We will develop a large language model (LLM) specialized in Japanese in collaboration with local corporations and expand into the Japanese market by partnering with local corporations."
Upstage Japan is the first overseas corporation established by Upstage in the Asian market. It is the second overseas corporation following the establishment of a U.S. entity in March of last year. Upstage is an AI solution startup founded by Kim, who oversaw the development of the LLM model 'Clova' as the leader of the Naver AI team in 2020. The company's main products are the AI-based optical character recognition (OCR) document processing technology 'Document Pass (DP)' and the self-developed large language model (LLM) 'Solar.' In July of last year, the company secured a sales channel to promote Solar to corporations using Amazon Web Services (AWS) by signing a business agreement.
Kim noted, "It is rare globally for corporations to have developed a full-stack AI model encompassing everything from OCR to LLM themselves," adding, "Upstage will strengthen its localized strategies with DP, which can convert any type of document into AI training data, and Solar, which was developed through self-pretraining, and will expand these into the global market."
Upstage possesses technological competitiveness in the field of specialized small language models (sLLM) rather than general-purpose LLM. Kim noted, "The version of 'Solar Pro 1.3' released this month has the highest benchmark performance among domestically developed models," adding, "In June, we plan to unveil the 'Solar Pro 1.5' version, expanded from the existing 22 billion parameters to 33 billion, and the first reasoning LLM model implementing 'Chain of Thought (CoT)' that matches OpenAI's 'O series' and DeepSeek's 'R1.'"
Kim mentioned, "Currently, many global corporations are showing great interest in our company, and discussions about investment are underway to some extent," adding, "We will soon have good news regarding investment attraction." So far, Upstage has secured a cumulative investment of 140 billion won. It is reported that investments have come from SK Networks, KT, and SBVA (previously SoftBank Ventures).
Kim also shared his thoughts on data copyright. He explained that he is pursuing collaboration with local corporations to secure data overseas. Kim said, "It has recently come to light that OpenAI and DeepSeek used data without authorization, and I believe that data use should be legal as much as possible," adding, "We have signed a business agreement with the U.S. AI chatbot company 'Quora' to utilize data, and in Korea, we are collaborating with various domestic corporations to use data."
Kim mentioned the potential for collaboration with AWS, which is preparing to enter the domestic public market. He stated, "We are discussing business related to the public sector with AWS, but we are not bound to AWS for business operations, so we can collaborate with other companies in the same way."