LG AI Research Unveils Next-Generation Hybrid AI Model EXAONE 4.0

LG AI Research has unveiled its next-generation hybrid AI model EXAONE 4.0, setting a new standard for global AI innovation. EXAONE 4.0 is a model that combines general natural language processing capabilities with the advanced reasoning capabilities proven through EXAONE Deep.

This model inherits a step-by-step, thought-based problem-solving approach, delivering outstanding performance in high-difficulty fields such as mathematics, science, and coding. It also supports Model Context Protocol (MCP) and Function Calling, laying the technical foundation for implementing Agentic AI, and has expanded its language support beyond Korean and English to include Spanish.

EXAONE 4.0 is offered in two lineups depending on use case.

• Professional model (32B): Suited for high-difficulty professional fields such as law, accounting, and medicine, it has demonstrated expertise sufficient to pass 6 nationally certified qualification exams.
• On-device model (1.2B): Compared to the previous model, its size has been halved while performance has improved, and it operates without a server connection, offering strengths in security and privacy protection.

For performance verification, the latest global benchmarks such as MMLU-Pro (81.8), LiveCodeBench v6 (66.7), GPQA-Diamond (75.4), and AIME 2025 (85.3) were used, demonstrating superior competitiveness compared to leading global models. LG AI Research has currently released EXAONE 4.0 via Hugging Face for research and academic purposes.