LG Group is pouring its efforts into using artificial intelligence (AI) technology, which Chairman Koo Kwang-mo of LG Group has emphasized as a future growth engine, to innovate customer experience and deliver new value. To that end, LG Group is accelerating investment and research centered on LG AI Research, the group's AI think tank.
According to the business community on the 29th, Koo said when establishing LG AI Research in 2020, "I will support and add strength so that LG AI Research can develop into the center of the global AI ecosystem, as the best talent and partners gather to freely take on the world's tough challenges."
LG AI Research is focusing on its proprietary AI model, EXAONE (LG AI Research), and on AI application research and development, helping affiliates solve tough problems while expanding the AI ecosystem through increased collaboration across different industries.
◇ LG Group's in-house AI EXAONE zeroes in on becoming the "world's best," beyond No. 1 in Korea
EXAONE means "EXpert AI for everyONE." EXAONE aims to be an "expert AI" that solves tough problems on industrial sites based on the industrial data of LG Group affiliates.
Koo said, "If we are confident in the direction, it is well worth investing," and is known to have actively supported LG AI Research in embarking on EXAONE development.
In December 2021, just a year after its launch, LG AI Research succeeded in developing EXAONE 1.0, Korea's first Multimodal AI model capable of processing various types of data such as text and images. Then in July 2023, LG AI Research developed EXAONE 2.0, a Multimodal AI trained on about 45 million pieces of professional literature and 350 million images. In Aug. 2024, it began releasing AI models externally by unveiling EXAONE 3.0 as an open-weights model on Hugging Face, a global open-source AI platform, for research and educational purposes.
The EXAONE 3.5 model unveiled by LG AI Research in Dec. 2024 achieved the feat of being the only Korean AI model selected as a "notable AI model" in a report published last year by the Stanford University Institute for Human-Centered Artificial Intelligence (HAI) in the United States.
In March and July 2025, LG AI Research also unveiled EXAONE Deep, Korea's first reasoning AI model, and EXAONE 4.0, a hybrid AI model that integrates general and reasoning capabilities. These models also made the list of "notable AI models," showcasing LG's technological prowess to the world. EXAONE 4.0 is the third hybrid AI model after those from Anthropic in the United States and Alibaba in China.
In November last year, Microsoft's (MS) AI Diffusion Report, which compared representative AI models by country, selected EXAONE 4.0 as Korea's representative model, placing it third after the United States and China. EXAONE 4.0 also set the fastest record among domestic AI models by surpassing 500,000 downloads within two weeks of its release.
In December last year, LG AI Research unveiled K-EXAONE, a frontier-class model with the world's best performance. In the government's first evaluation benchmark average scores, K-EXAONE showed 104% of the performance of Alibaba's similarly sized open-weights model Qwen3 235B (235 billion parameters) among the latest from the United States and China, and 103% of OpenAI's GPT-OSS 120B.
Building on five years of accumulated AI foundation model development expertise, LG AI Research completed K-EXAONE in five months and plans to develop a model that can compete head-on with the top-tier models of global big tech companies that have parameter counts in the trillions.
In April this year, LG AI Research unveiled EXAONE 4.5, a Vision-Language Model (VLM), once again proving LG's technological leadership that first developed Multimodal AI in Korea in 2021. EXAONE 4.5 excels at accurately reading and reasoning over complex documents actually handled in industrial settings, such as contracts, technical drawings, financial statements, and scanned documents.
EXAONE 4.5 received higher scores than GPT5-mini, Claude Sonnet 4.5, and Qwen3-VL in average scores across 13 benchmarks, including general visual understanding, infographics, and document understanding and reasoning metrics that read complex information in professional literature.
An LG Group official said, "LG AI Research aims to evolve EXAONE into physical intelligence that understands and judges the physical world beyond virtual environments."
◇ LG Group speeds up building "expert AI" to solve industrial challenges
LG is creating "expert AI" that can be used across industrial domains with global partners and affiliates.
LG AI Research developed EXAONE Business Intelligence (EXAONE-BI), a financial AI agent that predicts the direction of returns on investment assets and generates reports to aid decision-making by investors worldwide, based on unstructured data such as data from the London Stock Exchange Group (LSEG), news, and disclosures.
The collaboration between the two companies marks the first Korea-U.K. partnership in AI for finance, and on this basis, Kiwoom Securities decided to adopt EXAONE-BI in April this year, signaling the spread of Korea's AI technology in the global market.
LG affiliates are also delivering tangible results with AI technology across various business areas. LG Electronics will equip the 2026 Gram with On-device AI to provide document summarization, search, and translation via the EXAONE-based "Gram Chat On-device" service.
LG Display developed Hi-D, an EXAONE-based AI assistant for employees, improving work productivity with features such as real-time translation in video conferences. It also trained "AI production systems" with OLED manufacturing process domain knowledge to detect anomalies in real time during operations, identify causes, and take action, generating more than 200 billion won in annual cost effects based on process improvements.
LG Chem has actively adopted EXAONE Chemical Agent. EXAONE Chemical Agent is a technology that automates candidate discovery by having AI select candidate substances and predict synthesizability, and it can be applied to various fields such as new drug development and chemical substances to create synergy.
LG is also focusing on the convergence of AI and bio. EXAONE PASS 2.5 is a precision medical AI model that can reduce disease diagnosis time from two weeks to under one minute and has raised the accuracy of predicting gene mutations for major cancers to a world-class 76.75%. EXAONE PASS can be used as an "AI biomarker" that analyzes a patient's pathology images and recommends using target therapy A when there is a high probability of a specific gene mutation.
To realize "medical AI" that conquers cancer, LG AI Research is also collaborating with Professor Hwang Tae-hyun's team at Vanderbilt University Medical Center in the United States. The two organizations presented their "cancer agentic AI" research results at AACR 2026, the world's largest cancer conference, held in San Diego this year. The core of cancer agentic AI is that it is designed to perform the entire process from analyzing cancer patients' tissues to designing treatment strategies in a single day. LG AI Research and Professor Hwang's team plan to expand the scope of agentic AI applications from gastric cancer to various cancers, including colorectal and lung cancer.
LG AI Research is also developing "next-generation protein structure prediction AI" with Professor Baek Min-kyung's team at Seoul National University. LG expects that convergence between AI and bio will innovate customer experience and become an important catalyst for results in ABC (AI, Bio, Cleantech), the group's future growth engines.
Koo said, "We will take on the future where we can stay longer with our loved ones through innovative new drugs that treat intractable diseases," signaling the will to nurture the bio business as one of the future growth engines.