On 15 July, LG AI Research—the AI R&D arm of South Korea’s LG Group—unveiled Exaone 4.0, a groundbreaking hybrid reasoning AI model that combines general language processing with advanced reasoning capabilities, significantly expanding upon the strengths of its predecessor, the Exaone Deep model.

Exaone 4.0 is designed to excel in critical industry benchmarks across science, mathematics, and coding, outperforming well-known models from major tech firms such as Alibaba and Microsoft. However, it is worth noting that while Exaone has made strides, it still lags behind Deepseek’s top performer, illustrating the competitive nature of AI development.

What sets LG AI apart from its competitors is its targeted audience. Rather than pursuing the typical consumer market, LG has chosen to focus on business-to-business (B2B) solutions. This strategic decision aligns with the insights of Honglak Lee, who serves as co-head of LG AI Research and brings a wealth of experience as a former research scientist at Google Brain. Since its inception in December 2020, LG AI Research has centered on equipping enterprises with cutting-edge AI tools to enhance productivity and operational efficiency.

The introduction of Exaone 4.0 signals LG’s commitment to providing highly specialized AI solutions. For academic and research purposes, the model is now available on Hugging Face, a global open-source AI platform, enabling researchers and developers to experiment with its capabilities. Additionally, the latest version supports the Spanish language, thereby welcoming a broader audience and enhancing its usability outside of its original Korean and English competencies.

Expanding the Exaone Ecosystem

LG’s ambitions for Exaone extend beyond just a single model. Following the launch of Exaone 4.0, at the AI Talk 2025 event held on 22 July, the company revealed its expansive Exaone ecosystem and strategic roadmap aimed at a more comprehensive integration of AI technologies in enterprise solutions. Among the new offerings are Exaone 4.0 Vision Language—a multimodal AI model designed to interpret textual and visual data—and Exaone Path 2.0, focused primarily on healthcare applications that can diagnose patient conditions within minutes.

In addition to these models, LG has introduced enterprise-specific AI agents. One notable example is ChatExaone, currently being leveraged internally to optimize corporate workflows for LG employees. Moreover, Exaone Data Foundry is poised to revolutionize data generation, capable of achieving in a single day what typically takes a team of 60 experts three months to complete. This dramatic increase in efficiency underscores LG’s commitment to driving innovation in the AI space.

Performance and Efficiency Enhancements

Performance benchmarks demonstrate that Exaone 4.0 VL, set for release soon, surpasses Meta’s Llama 4 Scout in key performance areas, establishing LG AI as a formidable player in the AI arena. The research also highlights a partnership with FuriosaAI, a South Korean startup creating neural processing units (NPUs) specifically designed for machine learning tasks. The incorporation of FuriosaAI’s RNGD accelerators results in a performance boost of 2.25 times compared to competing GPUs, which positions LG’s model as a power-efficient alternative for enterprise-level applications.

By utilizing RNGD chips, a single rack can achieve up to 3.75 times more token generation for Exaone models than a traditional GPU rack operating under similar power constraints, suggesting not just a leap in computational capability but also a significant advancement in energy efficiency.

Future Implications and Security Applications

LG AI Research’s ultimate vision extends into the realm of autonomous agents aimed at enhancing enterprise security, illustrating the company’s recognition of AI’s dual role in fostering innovation while safeguarding sensitive information. This multifaceted approach to AI solutions positions LG as not just another player in the crowded AI landscape but as a specialized provider aimed at empowering enterprises with tailored, effective tools.

As LG AI Research continues to develop its Exaone ecosystem, the industry can expect to see substantial advancements that will redefine how businesses leverage artificial intelligence. With a clear commitment to B2B solutions and a strategic focus on enhancing productivity and security, LG is on the path to becoming a key leader in AI integration within the enterprise sector.