LG AI Research

LG AI Research is the artificial intelligence research subsidiary of LG Corporation, the South Korean industrial conglomerate that includes LG Electronics, LG Chem, LG Display, and other businesses. The subsidiary is headquartered in Seoul and develops the EXAONE family of large language models and the K-EXAONE Korean sovereign foundation model. LG AI Research's portfolio combines closed-weights commercial models distributed to LG group companies with open-weights releases through Hugging Face, and the lab has been consistently characterized in industry coverage as one of the leading Korean AI research organizations alongside Naver Cloud, SK Telecom, and Upstage.

At a glance

• Founded: December 2020 as a dedicated AI research subsidiary of LG Corporation, with research origins in LG's broader R&D organization dating to 2017.
• Status: Subsidiary of LG Corporation, the publicly traded South Korean industrial conglomerate. LG Corporation is listed on the Korea Exchange (003550.KS).
• Funding: LG Corporation R&D budget. Specific figures for the AI Research subsidiary are not separately disclosed from the broader corporate reporting.
• CEO: Lim Woo-hyung (Chief AI Officer at LG Corporation; head of LG AI Research; promoted in the December 2025 organizational changes as one of LG's youngest executive promotions).
• Other notable leadership: Honglak Lee (Senior Vice President, LG AI Research; concurrent appointment as Professor at the University of Michigan; senior research lead). Senior LG Corporation leadership including the Chairman Koo Kwang-mo (Chairman) provides strategic oversight.
• Open weights: Yes, partial. The EXAONE family has been released open-weights through Hugging Face under the LGAI-EXAONE organization. K-EXAONE and certain commercial-grade releases are closed-weights.
• Flagship models: EXAONE 4.5 (multimodal text-and-image reasoning), K-EXAONE (236-billion-parameter mixture-of-experts model with 23 billion active parameters; Korean sovereign foundation model), EXAONE 3.x and 4.0 series open-weights releases.

Origins

LG AI Research was established as a dedicated subsidiary in December 2020, consolidating LG Corporation's earlier AI research activities into a single research organization. The subsidiary's research origins date to LG Electronics and LG CNS investments in machine learning and AI research from approximately 2017, building on LG's longer-standing R&D capabilities in display, electronics, materials, and chemicals.

The first EXAONE model launched in December 2021 as Korea's first multimodal AI model. Through 2022 and 2023, LG AI Research expanded the EXAONE family across model scales and modalities, with research outputs covering text, image, and emerging multimodal capabilities. EXAONE was positioned for deployment across LG Corporation's industrial business lines, with applications spanning materials research, electronics product development, manufacturing optimization, and consumer-facing AI features in LG products.

The 2024 to 2025 release sequence expanded EXAONE significantly. The EXAONE 3.0 family included open-weights releases at multiple parameter scales, distributed through Hugging Face under the LGAI-EXAONE organization. The K-EXAONE 236-billion-parameter mixture-of-experts model with 23 billion active parameters launched in late 2024 as the Korean sovereign foundation-model entry, positioned as the Korean answer to the leading Chinese open-weights releases. Industry coverage characterized K-EXAONE as the first Korean model to enter the global top-10 capability ranking by certain measurements.

The EXAONE 4.5 release in 2026 introduced multimodal reasoning across text and images, with LG-internal benchmark results placing the model ahead of OpenAI's GPT-5 mini and Anthropic's Claude Sonnet 4.5 on a five-metric visual-reasoning evaluation. The release reinforced LG AI Research's positioning at the top tier of Korean AI capability development.

In April 2026, LG AI Research announced an expanded strategic alliance with NVIDIA, with LG AI Research head Lim Woo-hyung meeting NVIDIA Vice President Bryan Catanzaro to discuss integration of EXAONE with NVIDIA's Nemotron platform for sovereign-AI deployment. Korea's broader sovereign-AI policy framework, the Proprietary AI Foundation Model project overseen by the Ministry of Science and ICT, has provided approximately KRW 530 billion (approximately $358 million) of public-private funding for globally competitive Korean AI models by 2027, with K-EXAONE one of the principal funded projects.

The December 2025 LG Corporation organizational changes elevated Lim Woo-hyung as Chief AI Officer with expanded responsibility for the AI program across LG group companies. The reorganization signaled a structural step-up of AI as a strategic priority across the LG industrial conglomerate.

Mission and strategy

LG AI Research's stated mission, articulated through LG Corporation announcements and the EXAONE platform communications, is to provide artificial intelligence capability across LG group companies and to position EXAONE as a sovereign-AI alternative for Korean enterprise and public-sector customers. The strategy combines internal-deployment value creation across LG businesses with external-distribution participation in the Korean sovereign-AI ecosystem.

The strategy combines four threads. First, foundation-model research with the EXAONE family, covering text, multimodal, and reasoning capabilities. Second, internal deployment across LG group industrial businesses including LG Electronics consumer products, LG Chem materials research, LG Display, and LG Energy Solution battery research. Third, sovereign-AI participation in the Korean Ministry of Science and ICT's Proprietary AI Foundation Model project through K-EXAONE. Fourth, open-weights releases through Hugging Face for developer-community engagement and ecosystem building.

The competitive premise is that LG's industrial-application breadth provides a structurally distinct AI deployment opportunity that pure-play AI labs cannot match. The internal LG-business deployment is supported by AI capability tuned to industrial workflows, materials research, manufacturing optimization, and consumer-product integration, with external sovereign-AI participation as a complementary commercial channel.

The April 2026 NVIDIA strategic alliance positions LG AI Research as a key Korean partner in NVIDIA's broader Asia-Pacific sovereign-AI engagement, with EXAONE integrated into NVIDIA's Nemotron platform. The alliance combines LG's Korean-language and industrial-AI capabilities with NVIDIA's hardware-and-software ecosystem reach.

Models and products

• EXAONE 4.5. Released 2026. Multimodal AI model with simultaneous text-and-image reasoning capability. Reported leading position on internal LG visual-reasoning benchmarks against GPT-5 mini and Claude Sonnet 4.5.
• K-EXAONE. 236-billion-parameter mixture-of-experts model with 23 billion active parameters. Released late 2024 as Korea's principal sovereign foundation-model entry. Reported global top-10 capability ranking position at release.
• EXAONE 4.0 32B. Open-weights model released 2025. Reasoning-capable variant in the EXAONE family.
• EXAONE 3.x family. Earlier open-weights releases through Hugging Face, with multiple parameter scales for different deployment contexts.
• EXAONE Path. Vertically focused EXAONE variants for specific industrial application contexts.
• EXAONE Universe / EXAONE Atelier. Earlier specialized EXAONE applications for industrial and content-generation use cases.

The principal distribution channels are internal deployment across LG group companies, the EXAONE Studio platform for enterprise customers, the K-EXAONE consortium for Korean sovereign-AI deployment (including Elice and other partners through the Korean Ministry of Science and ICT framework), and Hugging Face for open-weights releases.

Benchmarks and standing

EXAONE 4.5's release in 2026 reported leading performance on LG-internal multimodal-reasoning benchmarks, with average score of 77.3 across five visual-reasoning metrics, ahead of GPT-5 mini at 73.5 and Claude Sonnet 4.5 at 74.6. Independent third-party benchmark coverage of the EXAONE 4.5 claims has been more limited.

K-EXAONE entered the global top-10 capability ranking at seventh position at its late-2024 release, the only Korean model in that top tier. EXAONE 4.0 32B has been a Korean reasoning-model leader, though the 2026 Naver Cloud HyperCLOVA X SEED Think 32B release reportedly took the leading Korean position on the Artificial Analysis Intelligence Index.

The EXAONE family's standing in the global AI landscape rests on the parent LG Corporation's industrial-application deployment scale, the K-EXAONE Korean sovereign-foundation-model positioning, the open-weights distribution through Hugging Face, and the NVIDIA strategic alliance. The benchmark profile is competitive at the Korean and broader Asian frontier-model tier.

Leadership

As of April 2026, LG AI Research's senior leadership includes:

• Lim Woo-hyung, Chief AI Officer at LG Corporation and head of LG AI Research. Promoted to expanded responsibility in the December 2025 LG Corporation organizational changes. Public face for LG's AI strategy and the EXAONE platform direction.
• Honglak Lee, Senior Vice President, LG AI Research. Concurrent Professor at the University of Michigan. Senior research-lead with academic-research credentials including machine-learning research output in deep generative models, reinforcement learning, and unsupervised learning.

LG Corporation Chairman Koo Kwang-mo provides strategic oversight for the AI program at the group-company level. The 2025 to 2026 organizational changes elevated AI as a strategic priority across the LG group, with continued senior-leadership expansion expected.

Funding and backers

LG Corporation's capital structure is the publicly listed parent (003550.KS), with LG Electronics (066570.KS), LG Chem (051910.KS), LG Display (034220.KS), LG Energy Solution (373220.KS), and other listed subsidiaries. LG AI Research operates as a dedicated subsidiary funded through LG Corporation R&D budget allocation.

Specific AI-research funding figures for the LG AI Research subsidiary are not separately disclosed from the broader LG Corporation R&D reporting. The Korean Ministry of Science and ICT's Proprietary AI Foundation Model project provides approximately KRW 530 billion ($358 million) of public-private funding through 2027, with K-EXAONE among the principal funded efforts.

The April 2026 NVIDIA strategic alliance includes joint development of next-generation AI models, with the alliance structure and any equity or compute-commitment terms not fully publicly disclosed.

Industry position

LG AI Research occupies a structurally distinctive position in the Korean and broader Asian AI landscape. The combination of the parent LG Corporation's industrial-application breadth, the EXAONE family's research depth, the K-EXAONE Korean sovereign-foundation-model positioning, the open-weights distribution through Hugging Face, and the NVIDIA strategic alliance produces a profile that differentiates LG AI Research from peer Korean AI organizations.

Industry coverage in 2025 and 2026 has frequently characterized LG AI Research as one of the four principal Korean AI research organizations alongside Naver Cloud, SK Telecom, and Upstage. The internal-LG-business deployment provides a use-case advantage that pure-play AI labs cannot match, particularly for industrial-AI applications in materials, manufacturing, electronics, and battery research.

Strategic risks include intensifying competition from Naver Cloud (the principal Korean AI peer), from US frontier labs operating in Korea through cloud distribution, and from Chinese open-weights releases. Strategic strengths include the LG industrial-application deployment scale, the EXAONE research depth, the K-EXAONE Korean sovereign-AI positioning, and the NVIDIA strategic relationship.

Competitive landscape

LG AI Research competes with and collaborates with several AI organizations:

• Naver Cloud. Direct Korean competitor through the HyperCLOVA X family. Naver Cloud's HyperCLOVA X SEED Think 32B reportedly took the leading Korean position on the Artificial Analysis Intelligence Index in early 2026, ahead of EXAONE 4.0 32B.
• SK Telecom. Direct Korean telecommunications-and-AI competitor through SK T-Brain.
• Upstage. Korean AI Insurgent through the Solar model family.
• Kakao Brain. Direct competitor as the AI division of Kakao.
• OpenAI, Anthropic, Google DeepMind, Microsoft AI. US frontier-and-incumbent competitors operating in Korea through cloud distribution.
• NVIDIA. Strategic partner through the April 2026 alliance and concurrent supplier of compute infrastructure.
• Aleph Alpha, Mistral AI, AI Singapore, Sarvam AI, Krutrim. Peer national-and-regional sovereign-AI organizations.

Outlook

Several open questions affect LG AI Research's trajectory in 2026 and 2027:

• The EXAONE 5.0 or successor release timing and capability profile.
• The K-EXAONE consortium and Korean sovereign-AI procurement dynamics under the Ministry of Science and ICT framework.
• The NVIDIA strategic alliance progression and any Nemotron-EXAONE integration outputs.
• Continued internal deployment across LG group industrial businesses and the impact on LG product lines and operations.
• The competitive dynamic with Naver Cloud, particularly following HyperCLOVA X SEED Think 32B's reasoning-leadership position.
• Continued senior-talent recruitment from Korean universities and from international AI organizations.
• Open-weights release strategy evolution including potential EXAONE 4.5 or successor open-weights variants.

Sources

• LG AI Research official site. Models, research, and capabilities reference.
• PR Newswire: LG Reveals Next-Gen Multimodal AI 'EXAONE 4.5'. EXAONE 4.5 release announcement.
• Korea Herald: LG says new AI beats GPT-5 mini in visual tests. EXAONE 4.5 capability coverage.
• Korea Herald: LG's K-Exaone breaks into global top 10 AI rankings. K-EXAONE release coverage.
• Korea Herald: LG, Nvidia deepen ties to scale AI ecosystem push. April 2026 NVIDIA alliance.
• LG: Organizational Changes for 2026. December 2025 leadership changes.
• KoreaTechDesk: Korea Moves Sovereign AI From Model Race to Deployment, Elice Joins LG Consortium to Operationalize K-EXAONE. K-EXAONE consortium context.