NC AI on Thursday unveiled VAETKI, a scalable multimodal generative foundation model aimed at supporting AI transformation in South Korea’s key industries. The model was trained on 10 trillion tokens and allocates 20 percent of its tokenizer vocabulary to Korean, including the ability to handle archaic language. NC AI said internal tests show stronger Korean benchmarks than GPT-OSS-120B series models and higher performance than Meta’s Llama 4 Scout on instruction following and reasoning. It also cuts KV cache memory use by about 83 percent.

Musinsa said on Jan. 8 it will open its first offline shoe specialty select shop, Musinsa Kicks, near Hongdae Entrance in Seoul on Jan. 9. The Hongdae store spans three floors and about 1,124 square metres, offering themed curation and a large shoe wall. It applies an O4O service using product QR codes and features category zones such as running and leather shoes. Musinsa plans additional stores in Seoul, including Seongsu and Gangnam, in the first half of this year.