LG has unveiled its new AI model, Exaone 4.0, which is noted as South Korea’s first hybrid reasoning AI system. The model can generate and evaluate hypotheses.
New AI Exaone 4.0
On Tuesday, LG announced the launch of the new AI Exaone 4.0. The company emphasized that this model is hybrid and capable of reasoning. Lee Jin Sik, head of the EXAONE Lab, said, "We will continue our R&D efforts to establish EXAONE as Korea’s leading frontier AI model and to demonstrate its competitiveness in the global market."
Development of Exaone at LG AI Research
LG AI Research was established in 2020 to accelerate the practical use of generative AI. The first model, Exaone 1.0, was launched in 2021. In March 2025, the company launched South Korea’s first reasoning AI model, Exaone Deep, featuring 32 billion parameters. This model demonstrates capabilities on par with Deepseek's R1 model, which has 671 billion parameters.
Comparison with Competitors
Exaone Deep has shown high results in various tests, outperforming Deepseek R1 and Alibaba QwQ-32B models. For instance, in the 2024 US Mathematical Olympiad, Exaone Deep-32B scored 90, while Deepseek and Alibaba scored 86.7. In Korea, the model also achieved impressive scores, obtaining 94.5 in the 2025 CSAT math exam. However, Exaone Deep still falls short in coding and language tasks compared to its competitors.
With the release of Exaone 4.0, LG is making strides in the development of reasoning AI and carving out its niche among local and international competitors. The model will be available in Seoul on July 22.
