LG
publicLG AI Research is the centralized AI hub of the LG Group, established in December 2020 to drive digital transformation across the conglomerate's electronics, chemicals, energy, and telecommunications divisions. Led by Chief Scientist Honglak Lee (former Google Brain researcher, University of Michigan professor), the lab focuses on "Expert AI" — models designed for high-stakes industrial, scientific, and professional tasks rather than general-purpose chat.
The lab's primary output is the EXAONE (Expert AI for Everyone) model family, which evolved from a 300B multimodal model (2021) through open-weight releases (EXAONE 3.0, 7.8B, 2024) to hybrid-attention architectures (EXAONE 4.0, 2025) and the current MoE flagship K-EXAONE (236B total / 23B active, with Multi-Token Prediction for 1.5x faster generation). EXAONE 4.0 32B achieved the highest Artificial Analysis Intelligence Index score of any 32B model.
Beyond language models, LG AI Research has produced notable work in materials science (GATE, screening billions of candidate molecules for data center cooling fluids), computational pathology (EXAONE Path 2.5, multi-omics alignment for cancer detection), and financial AI (EXAONE Business Intelligence with LSEG, powering an AI-driven ETF on NYSE). The lab is organized into specialized units: Superintelligence, Materials Intelligence, Physical Intelligence, and Bio Intelligence.
People
- Honglak Lee Google ScholarOpenReview — Chief Scientist & EVP (formerly Google Brain; Professor, University of Michigan)
- Woo-Hyung Lim — Head of LG AI Research
- Moontae Lee Google ScholarOpenReview — Head, Superintelligence Lab (formerly Cornell (PhD), Stanford (MS), Microsoft Research)
- Sehui Han — Head, Materials Intelligence Lab
- Seunghwan Kim — Head, Physical Intelligence Lab
News
- 2026-04-09 EXAONE 4.5 Released — LG's First Open-Weight Vision-Language Model — Korea Herald