The public evaluation of South Korea's independent AI foundation model project concluded on the 11th, moving the second evaluation into its final stage. The Ministry of Science and ICT plans to announce final results this month, reflecting scores from a public evaluation panel of 200 ordinary citizens. Naver Cloud and NC AI were eliminated in the first evaluation; in the second, one of four teams—LG AI Research, Upstage, SK Telecom, and Motif Technologies—will be dropped. On the 12th, ChosunBiz analyzed the four teams' models released on Hugging Face. LG AI Research's "K-EXAONE 2.0" more than tripled its parameters from 236 billion to 750 billion, the largest in Korea. SKT unveiled a 688 billion-parameter model, up from 519 billion. Motif's parameters total 314 billion, and Upstage's total 250 billion. All four models use a mixture-of-experts structure; active parameters are 37 billion for LG, 33 billion for SKT, 15 billion for Upstage, and 13.2 billion for Motif. Upstage and Motif pursue efficiency-focused strategies to curb computation per token, reducing GPU needs and costs. LG AI Research stated Korean researchers independently completed the entire process, from designing a 75-billion-class model to training and inference setup. SKT plans to scale its successor to a trillion-parameter-class model.