NC AI announced on the 16th that it has proven its Korean language-based AI technology by releasing four models as research open source, including the multimodal artificial intelligence (AI) model 'VARCO-VISION 2.0.'
The VARCO-VISION 2.0 14B model is characterized by its enhanced ability to analyze complex documents, tables, and charts simultaneously, as well as its improved bilingual processing capability in Korean and English. The 1.7B model is designed with a lightweight structure that can be utilized on personal devices such as smartphones and PCs.
Specialized in optical character recognition (OCR), VARCO-VISION-1.7B-OCR applies an AnyRes partitioning input structure, demonstrating high recognition accuracy even in mixed environments of Korean and English while processing high-resolution information. The VARCO-VISION embedding model calculates semantic similarity between text, images, and videos, enhancing search accuracy, and has confirmed excellent performance in the MultiVENT2.0 benchmark.
The models released this time are said by the company to have high practical applicability across various industries, including finance, education, culture, and manufacturing, for document automation, content creation, and video search. NC AI focused on enhancing cost efficiency and technical accessibility by maximizing data efficiency and designing lightweight models.
Lee Yeon-su, CEO of NC AI, noted, 'In the flow of the multimodal AI transition, VARCO-VISION 2.0 will serve as an opportunity to enhance the autonomy of domestic AI technology by combining Korean specialization with global-level technology.'