LG AI Research Adds EXAONE Models for Changing Factory Work

When factories change products or processes, the images used to inspect components can change too. LG AI Research says its new EXAONE expert models are designed to keep industrial AI working through those shifts while extending the lineup into scientific discovery and finance.

LG AI Research introduced the EXAONE-based models at its LG AI Talk Concert 2026. The release includes EXAONE Tabular for numerical table data, EXAONE Omni-Inspect for camera-image defect detection, EXAONE Discovery for science, and EXAONE Business Intelligence for finance.

Built to cope with changing conditions

Tabular is meant to find relationships in raw numerical tables and forecast future values. LG says it can adapt to new manufacturing environments with relatively small datasets. Omni-Inspect is designed to keep identifying component defects from camera images without retraining when products or processes change.

The jobs LG is assigning to the models

• Tabular handles numerical forecasting for manufacturing environments.
• Omni-Inspect spots component defects from camera images as factory products or processes change.
• Discovery screens scientific candidates, while Business Intelligence is the financial-sector offering.

A discovery claim and a planned lab loop

LG said Discovery, used with LG Household & Health Care, screened 420,000 candidate substances in one day and helped identify ramsidil, a substance that could help prevent hair loss. LG says the workflow compressed a process that typically takes 22 months into one day.

The institute also outlined a planned AI-powered autonomous laboratory. A materials model would predict synthesis outcomes, robotic equipment would run experiments, and the system would use the results to design the next experiments.

Specialization is the competitive case

The launch is not a claim that EXAONE leads the largest general-purpose models. At the event, Artificial Analysis co-founder George Cameron said Korean models including EXAONE trail leading U.S. and Chinese models in overall intelligence, but may remain competitive where cost, speed, flexibility and deployment choices matter more.

LG had already presented EXAONE’s industrial applications at ICML 2026 in July. This release gives that direction distinct models for specific industrial tasks, rather than treating one general model as the answer to every workflow.