LG AI Research Launches Korea’s Hybrid Exaone 4.0 Model

LG AI Research has unveiled Exaone 4.0, Korea's first hybrid AI model combining large language model fluency with advanced reasoning capabilities, designed to excel in specialized fields such as science, mathematics, biology, and chemistry. This model outperforms several leading AI models from Alibaba, Microsoft, and Mistral in benchmark tests, although it still trails behind some top-tier models like DeepSeek's R1. Exaone 4.0 supports multiple languages including Korean, English, and Spanish, and features Model Context Protocol technology enabling communication with external applications, paving the way for AI agents capable of complex tasks. The model consists of two versions: a 32 billion parameter expert model capable of passing six national professional license exams and a lighter 1.2 billion parameter on-device model optimized for consumer electronics, providing enhanced privacy and security by processing data locally. LG's partnership with FriendliAI now allows public deployment of Exaone 4.0 via serverless infrastructure, removing traditional barriers to access and making it available for research, education, and enterprise use. This development is part of LG's broader strategic investment in future technologies including AI, aiming to strengthen its competitive position in the global AI market.