← Sticker Wall MODEL · DETAIL

EXAONE Expert AI: Materials and New Drug Discovery Platform Subscription

1) Subscription and licensing fees charged to enterprises for EXAONE Discovery-type scientific discovery platforms; 2) p

MODEL

Key Fields

FIELD STAMPS

IndustryAI / LLM

RegionGlobal

ScaleGiant

ChannelOnline

📌 Background

In September 2026, LG AI Research unveiled its "Expert AI" strategy at the AI Talk Concert in Seoul, focusing on solving on-site industrial challenges, and announced it would expand external supply of EXAONE, which has been validated within the group. The multimodal EXAONE 4.5 released in April 2026 scored an average of 77.3 on five STEM benchmarks, and K-ExaOne was rated as the only Korean model to enter the global top ten, making research and industrial scenarios its differentiating selling point.

👤 Target Customers

R&D departments at chemical, battery, pharmaceutical, and materials companies, as well as large enterprises that need expert analysis of patents and academic papers

💰 Revenue Streams

1) Subscription and licensing fees charged to enterprises for EXAONE Discovery-type scientific discovery platforms; 2) project-based deployment fees for industry-customized expert models; 3) external supply of model APIs and private deployment licenses.

🧮 Cost Structure

Computing power for foundation model training and R&D talent are the largest costs, backed by LG Group investment; next are industry data acquisition, expert model fine-tuning, and customer delivery integration costs.

🛡️ Moat

LG Group's real industrial scenarios in chemicals, batteries, electronics, and other areas serve as a testing ground for the models, creating a data and trust loop of "validate internally, then sell externally"; Korean-language and specialized literature processing capabilities, plus STEM benchmark results, provide technical endorsement.

🔑 Keys to Success

Turn internally validated group cases into industry benchmarks that can be showcased externally
Establish quantifiable results in vertical scenarios such as materials and new drug discovery

⚠️ Risks

Internal scenario experience is difficult to replicate in external customer industries
General-purpose large-model vendors moving down into vertical scientific scenarios will squeeze space

🏢 Cases

In September 2026, LG AI Talk Concert unveiled Expert AI designed for on-site industrial challenges and announced expanded external supply of EXAONE
The multimodal EXAONE 4.5 model scored an average of 77.3 on five STEM benchmarks
K-ExaOne ranked seventh in a global AI performance evaluation, the only Korean model to enter the top ten

📊 SWOT Analysis

Strengths

Proprietary training and validation data from the group's internal chemical and biological scenarios
Expert AI positioning avoids head-on competition with general-purpose models

Weaknesses

Limited market size in Korea, and weaker overseas brand recognition than US models
Commercial revenue scale is still small, relying on parent company funding

Opportunities

Global manufacturing companies are seeking vertical scientific AI as an alternative to general-purpose models
Countries are promoting sovereign AI and local models, giving Korean models a policy window

Threats

Giants such as OpenAI and Google are also entering the scientific discovery track
Open-source models catching up in performance is compressing licensing pricing space

Sources