LG AI Research has launched Exaone 4.0, Korea's first hybrid AI model that combines large language understanding with advanced reasoning capabilities, building on its predecessor, Exaone 3.5.

The company plans to showcase Exaone 4.0 in Seoul on July 22, 2025, emphasizing its innovative features and performance.

LG has partnered with Friendly AI to offer a commercial API service for Exaone 4.0, making it more accessible for developers and businesses without requiring high-end hardware.

The 32-billion-parameter expert model achieved top scores in evaluations like 81.8 on MMLU Pro and 84.8 on IFEval, outperforming competitors such as DeepSeek and GPT-4o mini.

This expert model passed exams for six national professional certifications, including medical and dental licenses, demonstrating its high level of expertise.

An embedded on-device model with 1.2 billion parameters outperformed OpenAI's GPT-4o mini in fields like mathematics and coding, and is designed for use in devices such as smartphones and home appliances.

The on-device model operates independently of external servers, ensuring faster processing, enhanced privacy, and robust security, suitable for applications in smartphones, automotive systems, and robots.

Exaone Deep-32B scored highly in various benchmarks, including 94.5 in South Korea's CSAT math section, and outperformed leading hybrid AI models globally in assessments like knowledge, coding, science, and mathematics.

Only a few global companies, including Anthropic, Alibaba, and OpenAI, are developing similar hybrid AI models, with OpenAI working on GPT-5.

The AI landscape in South Korea is becoming increasingly competitive, with major players like Naver, SK Telecom, and Kakao developing models and forming U.S. partnerships.

LG's collaboration with partners aims to make generative AI models more accessible and cost-effective, accelerating product development.

LG emphasizes ongoing research and development to enhance Exaone 4.0's global competitiveness, aligning with the company's broader strategy to invest heavily in future technologies.