LG AI Research

What LG AI Research does

LG AI Research is the dedicated artificial intelligence research subsidiary of LG Corporation, established in 2020 and headquartered in Seoul, South Korea, with an additional research center in Ann Arbor, Michigan. The organization developed EXAONE — Korea's first multimodal AI model — and has since evolved it into EXAONE 4.5, a native multimodal foundation model that combines vision and language understanding from the design stage via an integrated Visual Encoder, with MTP (Main Token Prediction) acceleration delivering roughly 1.5x speed improvements over prior versions. Beyond the core foundation model, the company operates eight research domains (Superintelligence, EXAONE, Language, Physical Intelligence, Bio Intelligence, Data Intelligence, Materials Intelligence, and Advanced Agent) and ships a portfolio of products including ChatEXAONE (a document/web/data agent), EXAONE Data Foundry (LLM tuning), EXAONE Business Intelligence (forecasting/decision agents), EXAONE Chemical Agent (materials design), a Multi-Agent Reinforcement Learning-based Naphtha Cracking Center scheduler, and a Robot Foundation Model for cross-embodiment physical AI.

The company's business model is that of an internal R&D and solutions arm for LG Group affiliates — primarily LG Chem and LG Electronics — rather than an independent commercial software vendor. Revenue is generated indirectly via operational efficiency, scheduling optimization, and automation deployed within LG's industrial operations, with named deployments including the LG Chem Daesan NCC facility and a PoC with LG Electronics' Production Engineering Research Institute for manufacturing robotics. The company is also pursuing a product-led growth motion externally through the EXAONE Showroom self-serve platform and open distribution of EXAONE 4.5 on Hugging Face, though no public pricing, external enterprise logos, or independent revenue figures are disclosed. LG AI Research has no external funding rounds, no disclosed management team, and no public listing, indicating full ownership and funding by LG Corporation.

Name

LG AI Research

Legal name

LG AI Research

Website

https://lgresearch.ai

Company type

Private

Founded year

2020

Operating status

Operating

Headcount range

251–500 employees

Ownership category

akta.pro rank

Product category

AI Foundation Models and Industrial AI Solutions

NAICS

Computer Systems Design and Related Services (5415), Research and Development in the Physical, Engineering, and Life Sciences (except Nanotechnology and Biotechnology) (541715), Research and Development in Biotechnology (except Nanobiotechnology) (541714)

SIC

Services-Computer Programming Services (7371), Services-Commercial Physical & Biological Research (8731), Services-Prepackaged Software (7372)

akta.pro primary industry

Agents & Autonomous Workflows (Tool Use, Planning, Multi-Agent) (HDAAACAF)

akta.pro secondary industries

Open-Source Model Ecosystems & Model Marketplaces (HDAAACAM), LLMOps & Generative AI Platforms (Prompt/Agent Orchestration, RAG) (HDAEANAD), Enterprise Foundation Model Integration & APIs (Connectors, Governance, Deployment) (HDAAACAO)

Keywords

• Multimodal AI foundation models
• Industrial AI optimization
• Multi-agent reinforcement learning
• Robot foundation models
• LLM tuning platform

Headquarters

HQ city

Seoul

HQ country

South Korea

HQ region

Asia

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales

Revenue model

1. Internal R&D for LG Group: LG AI Research operates as the AI research and development arm of LG Group, developing AI technologies for internal deployment across LG affiliates including LG Chem and LG Electronics. Revenue is generated indirectly through operational efficiency and competitive advantages provided to parent company.

Go-to-market motion2 records

Distribution channels3 records

Marketing channels8 records

Core offering

LG AI Research develops and deploys the EXAONE family of AI foundation models — multimodal language-and-vision models built natively from the design stage — along with derivative enterprise applications such as conversational agents (ChatEXAONE), an LLM tuning platform (EXAONE Data Foundry), a decision-intelligence suite (EXAONE Business Intelligence), a materials-discovery agent (EXAONE Chemical Agent), a multi-agent reinforcement-learning optimizer for naphtha cracking operations, and a Robot Foundation Model for physical AI. Solutions are delivered via a self-serve web showroom, open model weights on Hugging Face, and direct enterprise deployments primarily with LG Group affiliates and industrial partners.

Product overview

LG AI Research offers EXAONE as its core multimodal AI foundation model ecosystem, developed since 2020 as Korea's first multimodal model. The portfolio includes EXAONE 4.5 (the latest version with native multimodal architecture), ChatEXAONE (conversational agent), EXAONE Data Foundry (LLM tuning platform), EXAONE Business Intelligence (forecasting/decision AI), EXAONE Chemical Agent (materials discovery), and Naphtha Cracking Center Scheduling Optimization (industrial MARL solution). The EXAONE Showroom serves as the web platform for accessing these solutions. Research areas span Superintelligence, Language, Physical Intelligence (including Robot Foundation Model), Bio Intelligence, Data Intelligence, Materials Intelligence, and Advanced Agent.

Differentiator

Problem solved

Functional benefit

Brands

• EXAONE: LG AI Research's flagship multimodal AI model. EXAONE 4.5 features native multimodal architecture combining vision and language, designed for complex industrial applications.
• EXAONE Data Foundry
• ChatEXAONE
• EXAONE Business Intelligence
• EXAONE Chemical Agent
• EXAONE Naphtha
• Robot Foundation Model (RFM)

Products and services

• EXAONE LG AI Research's flagship multimodal AI foundation model brand, launched in 2020 as Korea's first multimodal model and evolved through multiple versions to the current EXAONE 4.5 with native multimodal architecture combining vision and language. Sold/licensed to enterprise customers and developers, with weights also distributed via Hugging Face.
• EXAONE 4.5 Latest EXAONE foundation model version with native multimodal architecture featuring integrated Visual Encoder, ~1.5x speed improvement via MTP, multilingual vocabulary support, and superior professional document understanding and Korean visual context comprehension. Available to enterprise customers and developers via Hugging Face.
• ChatEXAONE Conversational AI agent that brings together documents, web, and data into one seamless interface, designed to operate as an expert assistant that enhances productivity and supports smarter enterprise decision-making. Available via the EXAONE Showroom.
• EXAONE Data Foundry End-to-end LLM tuning platform that generates domain-specific QA data from real-world documents and accelerates optimization through rapid feedback loops, enabling enterprises to customize EXAONE for their domains without ground-up training. Available via the EXAONE Showroom.
• EXAONE Business Intelligence Modular AI system that supports forecasting, interpretation, and decision-making in dynamic environments, powered by specialized agents for time-series analysis, causality, generative modeling, and decision intelligence. Available via the EXAONE Showroom.
• EXAONE Chemical Agent AI agent for designing target-optimized materials and predicting material properties without experiments, enabling faster, cheaper, and more successful discovery workflows for chemical and materials R&D organizations. Available via the EXAONE Showroom labs.
• Naphtha Cracking Center Scheduling Optimization Multi-Agent Reinforcement Learning solution for optimizing Naphtha Cracking Center operations across the receipt, blending, and furnace stages, deployed at the LG Chem Daesan facility via an online web service where plant staff obtain AI-generated schedules in real time.
• Robot Foundation Model (RFM) Universal robotic intelligence research initiative combining a Vision-Language-Action framework for robot action generation with a World Model approach for predicting future environmental states, targeting cross-embodiment robot control across industrial and domestic environments.
• EXAONE Showroom Web-based self-serve platform (showroom.exaone.ai) showcasing EXAONE solutions including Data Foundry, ChatEXAONE, Business Intelligence, Chemical Agent, and labs for molecule design and physical AI, enabling enterprise customers and developers to access and evaluate EXAONE capabilities.

Quantifiable outcome

• Naphtha Cracking Center AI scheduler shows high profit and constraint satisfaction in deployment
• +2 more outcomes

Named customers1 record

Segments3 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability13 records

Feature9 records

Partnerships

Two partnerships are on record, tiered core.

• LG ChemcoreStrategic or Co-development Partner · 1 January 2024LG AI Research collaborated with LG Chem to develop and deploy Multi-Agent Reinforcement Learning-based NCC scheduling optimization for the LG Chem Daesan Naphtha Cracking Center facility. Researchers from LG AI Research and LG Chem co-authored publications and developed an online web service allowing LG Chem staff to generate NCC schedules in real-time. Staff are now operating the facility based on AI-generated schedules.
• LG Electronics Production Engineering Research Institute (PRI)coreStrategic or Co-development Partner · 1 January 2024LG AI Research is collaborating with LG Electronics Production Engineering Research Institute (PRI) to deploy Robot Foundation Models and physical AI solutions into live manufacturing ecosystems. This Proof of Concept (PoC) initiative aims to validate whether RFMs can accurately digest on-site operational instructions and visual feeds to solve complex automation bottlenecks in factory environments.

Scale indicators3 records

Recent moves5 records

Expansion highlights6 records

Direct peers

• Naver Cloud (Hyperclova): Naver's Hyperclova is Korea's other flagship sovereign large language model, developed by Naver Cloud and Naver Labs. Directly comparable as a Korean foundation model competing for the same domestic enterprise and government customers that LG AI Research targets.
• Kakao (KoGPT / Kanana): Kakao has developed KoGPT/Kanana as a Korean-language foundation model and offers consumer-facing AI agents via KakaoTalk. Comparable as a Korean conglomerate-backed LLM developer pursuing enterprise and consumer agent products similar to EXAONE/ChatEXAONE.
• SK Telecom (A.X / A.X-Enterprise): SK Telecom's A.X series is a Korean foundation model co-developed with Anthropic, positioned for telecom, enterprise, and physical-AI applications. Comparable as a Korean carrier/conglomerate-backed foundation model pursuing adjacent industrial AI verticals.
• Upstage: Upstage develops Solar, a Korean-focused LLM family, and offers document-AI enterprise products. Directly comparable as an independent Korean foundation model vendor targeting enterprise document understanding and agent use cases similar to EXAONE Data Foundry and ChatEXAONE.
• Meta AI (Llama): Meta's Llama is the leading open-weight multimodal foundation model distributed on Hugging Face. Directly comparable as an open-source multimodal LLM competing with EXAONE 4.5 for the same enterprise fine-tuning and developer mindshare.
• Mistral AI: Mistral AI develops open-weight and commercial LLMs with strong enterprise traction in Europe. Comparable as an open-source-friendly foundation model vendor pursuing enterprise deployment and fine-tuning — overlapping directly with EXAONE's Hugging Face distribution strategy.
• Physical Intelligence: Physical Intelligence is a US-based startup building foundation models for robot control using VLA and large-scale robot data. Directly comparable to LG AI Research's Robot Foundation Model effort as a peer in the emerging robot foundation model category.

Broad incumbents

• Anthropic: Anthropic develops Claude, a frontier LLM with strong enterprise agent and reasoning capabilities, and is partnered with SK Telecom in Korea. Comparable as a broad incumbent foundation model provider whose enterprise agent offering overlaps with ChatEXAONE and EXAONE Business Intelligence.
• OpenAI: OpenAI is the dominant general-purpose foundation model and agent platform (GPT-4o, ChatGPT Enterprise). Comparable as a broad incumbent whose enterprise APIs and agent products compete with EXAONE-based offerings in the global enterprise market.

Others

• Hugging Face: Hugging Face hosts EXAONE 4.5 and is the de facto open-model marketplace and developer ecosystem that LG AI Research leverages for distribution. Relevant as an enabling infrastructure/ecosystem participant rather than a direct competitor.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat6 records

Key risks6 records

Key highlights6 records

Customer concentration

• YouTubeyoutube.com/channel/UCaoWx-ScnGYSrfKKcOn-YqA?view_as=subscriber
• Facebookfacebook.com/LG.ResearchAI
• LinkedInlinkedin.com/company/lgairesearch
• Xx.com/LG_AI_Research

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about LG AI Research