LG Group Launches 'One LG' Strategy for Physical AI with EXAONE 4.5

BARCELONA — LG Group is intensifying competition in physical AI, leveraging its proprietary artificial intelligence foundation model "EXAONE."

LG AI Research Institute, which is developing EXAONE, plans to advance next-generation vision language models (VLM) that will serve as the brain for humanoid robots. LG Uplus will build the Seoul metropolitan area's largest AI data center (AIDC) to provide full-stack AI services under the "One LG" strategy.

Lee Sang-yeop, Chief Technology Officer at LG Uplus, and Lim Woo-hyung, Co-Director of LG AI Research Institute, unveiled the "One Team LG" technology roadmap at a press briefing held in Barcelona, Spain, on March 1 local time.

LG plans to release its next-generation AI model "EXAONE 4.5" in the first half of this year. EXAONE 4.5 is a VLM that combines language intelligence with visual intelligence. Unlike conventional large language models (LLM) that process only text, the model features multimodal technology capable of comprehensively understanding and analyzing both text and images. The model will be installed as the brain of "KAPEX," the Korean humanoid robot LG is developing. The company plans to apply EXAONE 4.5 to KAPEX to enhance the robot's cognitive and decision-making functions.

LG aims to elevate "K-EXAONE" to the highest-performing language model among global open-weight models in the second round of evaluation for its proprietary AI foundation model project in the first half of this year. Open-weight refers to a method where the weights of a trained AI model are made public, allowing users to run or modify it for their purposes. LG AI Research Institute plans to release EXAONE 4.5 as an open-weight model soon.

The "One Team LG" strategy also applies to next-generation agentic AI development. The strategy aims to achieve optimal performance by organically combining LG AI Research Institute's AI models with LG Uplus's telecommunications technology capabilities. Through this approach, LG expects EXAONE to establish itself as an AI that operates stably and cost-effectively within enterprise environments.

Infrastructure to support this initiative is also being expanded. LG Uplus is building the Seoul metropolitan area's largest AIDC (200MW) in Paju, with completion targeted for 2027. The large-scale infrastructure, capable of accommodating up to 120,000 graphics processing units (GPU), is expected to serve as the hub for the "One LG" strategy, consolidating core capabilities of affiliates including LG Electronics, LG Energy Solution, and LG CNS.

"Competitiveness in the AI era is not simply about securing larger models or more computing resources," CTO Lee said. "The key battleground is how precisely we can design agentic architectures that evolve on their own through a cyclical structure of planning, execution, evaluation, and modification."

He added, "Through close collaboration with LG AI Research Institute, LG Uplus will set the standard for personalized agentic AI that becomes more sophisticated over time."

Director Lim also stated, "Based on development know-how accumulated over several years, we will introduce multimodal AI models on a different level. Beyond AX, we will focus on implementing AI that creates tangible value in the physical real world."