LG AI Research Institute on the 31st released the second model of the Ministry of Science and ICT-led "independent AI foundation model project," "K-EXAONE 2.0," to the open-source community Hugging Face.
K-EXAONE 2.0, with 750B (75 billion) parameters, the largest in Korea, is more than three times bigger than the first model (236B). LG AI Research Institute fully opened its model licensing policy to allow anyone to use it commercially.
K-EXAONE 2.0 scored an average of 70.1 across 24 benchmark evaluation metrics, achieving performance more than 10% higher than the first model (63.3). In particular, the average performance of three key indicators in coding and agentic coding rose 30%.
K-EXAONE 2.0's supported languages expanded to 10 in total, adding French, Italian, Portuguese, and Polish to the existing Korean, English, Spanish, German, Japanese, and Vietnamese. This is the highest number among AI foundation models in Korea.
K-EXAONE 2.0 also achieved an average score of 94.6 in evaluations for compliance with Korea's particularities and global ethics standards (KGC-Saftey) and in safety assessments related to geopolitical context judgment (ROK-Fortress), scoring higher than global leading models.
Im Woo-hyeong, co-head of research at LG AI Research Institute, said, "K-EXAONE 2.0 is meaningful in that domestic researchers independently completed the entire process for a 750B-class model, from design to data training, distributed training, and inference environment setup," and emphasized, "We have secured the capability to compete in the same weight class as global frontier models."
LG AI Research Institute plans to further strengthen its "expert AI" portfolio by unveiling a new industry-specialized AI foundation model next week.
During the second review of the "independent AI foundation model project," the development of a public evaluation site for real-world usability assessment is being finalized, and a service is being prepared that will allow the entire public to experience "K-EXAONE 2.0."