On the 22nd, the LG AI Research Institute held the 'LG AI Talk Concert 2025' at the Convergence Hall of LG Science Park in Magok, Gangseo-gu, Seoul, and announced the ecosystem of 'EXAONE', a large language model (LLM) developed over the past five years.
Lee Hong-rak, co-director of the LG AI Research Institute and Chief AI Scientist (CSAI), introduced the foundation model of LG AI Research Institute, starting with the recently unveiled 'EXAONE 4.0' and 'EXAONE PASS 2.0'.
'EXAONE 4.0' is the first hybrid AI model in Korea, and 'EXAONE PASS 2.0' is a precision medical AI model that can shorten disease diagnosis time from two weeks to within one minute. The multi-modal AI model 'EXAONE 4.0 VL (Vision Language)', which comprehends complex professional documents, images, and molecular structures, was also unveiled for the first time.
Following this, Choi Jeong-kyu, head of the AI Agent Group at LG AI Research Institute, unveiled sequentially the 'ChatEXAONE', an enterprise AI agent based on EXAONE that has completed internal validation, and 'EXAONE Data Foundry' and 'EXAONE On-Premise'.
Choi said, 'ChatEXAONE, the AI agent for LG employees, has obtained ISO certification, allowing it to use documents classified as national core technologies.' He also noted, 'Since the unveiling of EXAONE 4.0, we have expanded the model license scope to include educational purposes, and we are preparing to enable students from elementary school to university to freely utilize EXAONE.'
The LG AI Research Institute also introduced key technologies that can rapidly create specialized AI models. The EXAONE Data Foundry unveiled that day serves as AI infrastructure that produces high-quality data, allowing one person to complete what would typically require 60 experts working for three months in just one day.
The LG AI Research Institute explained that as a result of pilot projects conducted with LG affiliates and national agencies, data productivity has increased by at least 1,000 times, and data quality has improved by an average of over 20% compared to previous methods.
Introduced as the final part of the EXAONE ecosystem was 'EXAONE On-Premise', completed with purely domestic technology from AI semiconductors to models. EXAONE On-Premise is a solution that allows corporations to build enterprise agents in an environment independent from external threats, enabling the use of the EXAONE model without security concerns.
The LG AI Research Institute introduced key partners from both domestic and international markets in the EXAONE ecosystem and discussed the potential for expanding AI B2B business model diversification strategies.
Professor Baek Min-kyung from the Department of Life Sciences at Seoul National University, who is collaborating with the LG AI Research Institute, explained the preparation status of 'next-generation protein structure prediction AI' for drug development. Professor Baek stated that they are creating an AI model that surpasses Google DeepMind's AlphaFold in the field of disease treatment.
Arman Sahovitch, head of data platform solutions for the Asia-Pacific region at the London Stock Exchange Group, introduced a business intelligence service created with EXAONE. The LG AI Research Institute is preparing a service to predict the revenue direction of investment assets based on unstructured data like the data, news, and disclosures from the London Stock Exchange Group, aiding decision-making for investors worldwide.
Lim Woo-hyung, co-director of the LG AI Research Institute, remarked, 'We plan to secure global competitiveness based on our own foundation model and apply AI models to various industrial fields,' adding, 'We will create an AI ecosystem together with global partners.'