Home > Business > Tech

LG heightens AI ambitions with K-Exaone model and Korea's largest data center

Published: 02 Mar. 2026, 18:11 Updated: 02 Mar. 2026, 19:07

Lee Sang-yeob, the chief technology officer of LG U+, left, and Lim Woo-hyung, the co-head of LG AI Research, speak during a press conference for Mobile World Congress 2026 in Barcelona, Spain, on March 1. [LG U+]

BARCELONA, Spain — LG has escalated its AI ambitions by pledging to turn its proprietary K-Exaone model into a top-tier global open-weight system and backing it with what it says will be Korea’s largest AI data center.

With an open-weight system, developers will be able to access the model's trained parameters and build on them.

The institute also said Exaone 4.5, currently under development, will be unveiled in the first half of the year as a vision-language model designed to serve as the “brain” for humanoid robots and other physical AI systems.

At a joint press briefing with LG U+ during Mobile World Congress 2026 in Barcelona, Spain, on Sunday, LG AI Research outlined its road map.

Lim Woo-hyung, the co-head of LG AI Research, said K-Exaone ranked seventh among global open-weight models on the Artificial Analysis Intelligence Index, a global AI evaluation platform, in January.

“Rather than the ranking itself being significant, what matters is that we are now competing alongside leading global models in a constantly evolving sector,” Lim said. “The second phase of the foundation model project will be underway through the first half of this year, and our goal is to achieve the top performance among global open-weight models.”

The company also unveiled plans to link its top-performing AI models to large-scale infrastructure for industrial deployment. Central to that strategy is the AI data center in Paju, Gyeonggi, scheduled for completion next year.

Lee Sang-yeob, the chief technology officer of LG U+, said LG affiliates are preparing for what he described as Korea's largest AI data center in coordination with the anticipated completion of K-Exaone.

LG AI Research logo [LG AI RESEARCH]

“The Paju AI data center will have a capacity of 200 megawatts and will be able to accommodate up to 120,000 GPUs,” Lee said. “We aim to offer end-to-end services connecting AI data centers to AI models.”

LG AI Research also confirmed that Exaone 4.5 — a vision-language model that integrates language and visual intelligence — will be released in the first half of the year.

“Back in 2021, we developed Korea’s first multimodal AI model, Exaone 1.0,” Lim said. “Building on that expertise, we are now capable of fully developing key components, including vision encoders, in-house.”

LG U+ plans to upgrade its Exaone-based AI agent ixi-O and expand its applications to physical AI systems.

“The goal of ixi-O is to go beyond smartphones and become an ambient AI agent that connects a wide range of devices — including humanoid robots, vehicles and smart glasses — around us,” Lee said.

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
BY EO HWAN-HEE [[email protected]]