"Advance AI Models... Expand Their Range of Applications"

"Although we failed to advance in the second round of the government's evaluation of independent foundation models (IFMs), we will continue striving to develop Korea's top large language model (LLM)."


In a recent interview with The Asia Business Daily, Lim Jeonghwan, CEO of Motif Technologies, said, "Although we are disappointed that we were not selected for the IFM project, it was a good opportunity for us to grow as a startup and make our name known. We will continue developing AI models by securing additional graphics processing units (GPUs)."


Motif Technologies CEO Lim Jeonghwan speaks in an interview with The Asia Business Daily on September 21. Photo by Seo Sojeong.

Motif Technologies CEO Lim Jeonghwan speaks in an interview with The Asia Business Daily on September 21. Photo by Seo Sojeong.

View original image

Lim said, "In our technical report, we directly compared our performance with leading open-weight models from overseas and confirmed our competitiveness in model design and inference efficiency: we achieved comparable performance with a smaller architecture in terms of both total and active parameters. Unlike global Big Tech companies, which invest massive amounts of GPU resources and manpower, Motif achieved this in about five months with a team of just 30 and around 800 GPUs provided by the government."


Motif joined the IFM project led by the Ministry of Science and ICT as a late entrant. In the second-stage evaluation held in August, the company surprised observers by ranking first among Korean models in the Artificial Analysis Intelligence Index (AAII). Rather than fine-tuning overseas open-source models, Motif developed its model from scratch, designing the architecture itself. It rapidly improved the model's performance, placing ninth among LLMs globally and among the world's top three open-weight models, and drew attention from developers overseas. Despite receiving high scores in benchmark evaluations, Motif was ultimately not selected and formally appealed to the government.


Lim said, "We wanted to know specifically why we were ultimately not selected, and felt that the issue concerned the company's credibility and reputation. In the end, the government did not provide us with detailed reasons, including about the evaluation process. However, participating in the IFM project gave us an opportunity to demonstrate Motif's technology. We have also gained global recognition after attracting attention from tech influencers starting late last year, so we will continue our development efforts."


One change since joining the IFM project is that competing companies have begun taking an interest in Motif's technology. Competitors have approached Motif's key engineers with job offers, and more companies have requested collaboration. Motif has signed business agreements with organizations and companies in various fields, including education and energy, and is carrying out AI projects for businesses while building its capabilities in customized AI transformation (AX).


In September, Lim traveled to the United States and met with Nvidia representatives to discuss GPUs and other matters. Motif also decided to secure 1,000 GPUs independently through a cloud provider and use them to develop AI models. Lim said, "I believe it is a significant achievement for a Korean startup to rank among the world's leaders with its own technology in the frontier LLM market, which has been dominated by the United States and China. We have demonstrated that it is possible to develop competitive AI without relying on overseas models."



He added, "We will continue improving our models, developing next-generation models, and strengthening our technological competitiveness among the world's leaders. We will also expand the range of applications so that Motif's AI can be used in many settings, including everyday life and workplaces."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing