NC AI releases new multimodal AI model as open source

NCSOFT's artificial intelligence (AI) subsidiary NCAI demonstrates its "Avatar Shift" technology to create AI avatar animation at the world's largest telecommunications fair "MWC 2025" in Barcelona, Spain, March 11. Courtesy of NCSOFT

NC AI, an artificial intelligence (AI) affiliate of game giant NCSOFT, said Wednesday it has released its new Korean-based multimodal AI model as open source.

Its four models of VARCO-VISION 2.0 — 14B, 1.7B, 1.7B OCR and Video-Embedding — are now available for free for commercial purposes, according to NC AI.

VARCO-VISION 2.0 is a next-generation AI model capable of understanding both images and text simultaneously to answer questions. It can analyze multiple images at once, making it effective in processing complex documents, tables and charts. It also offers natural understanding of both Korean and English, with improved comprehension of Korean culture.

NC AI said the 14B model outperforms other open-source counterparts, including InternVL3-14B, Ovis2-16B and Qwen2.5-VL 7B, in various benchmarks, such as English and Korean image understanding and optical character recognition (OCR).

"With the release of these four models, NC AI, which has been leading Korea's multimodal AI market in specialized fields, such as media, games and fashion, has demonstrated its potential to uphold Korea's sovereignty in the vision-language AI domain," NC AI CEO Lee Yeon-soo said.