SEOUL, January 15 (AJP) - LG AI Research, SK Telecom and Upstage have passed the first evaluation round of South Korea’s competition-style national project to select two homogeneous artificial intelligence foundation models by the end of the year.
According to results announced Thursday by the Ministry of Science and ICT, three of the five contenders — Naver Cloud, NC AI, Upstage, SK Telecom and LG AI Research — advanced directly based on aggregate scores, while one additional bidder will be reinstated through a secondary evaluation. The rematch will be held between Naver Cloud and NC AI.
The ministry disclosed the results at a briefing at the Government Complex Seoul, chaired by Second Vice Minister Ryu Je-myung.
The government-led project was launched under the banner of “leaping into the world’s top three AI powers,” positioning it as a flagship pillar of South Korea’s sovereign AI strategy. The program provides state support for GPU resources, data access and engineering costs, with the goal of developing domestically controlled foundation models capable of achieving at least 95 percent of leading global AI benchmark performance.
Under the project guidelines, eligible models are defined as those designed and pre-trained domestically, excluding derivative systems fine-tuned from foreign base models.
The ministry explained that the first-round evaluation combined AI benchmark tests — quantitative measures of technical performance — with expert and user assessments to comprehensively evaluate model performance, real-world applicability, cost efficiency relative to model size, and potential spillover effects on the domestic and global AI ecosystem.
In the benchmark evaluation, LG AI Research recorded the highest score with 33.6 out of 40 points. The company also topped the expert review with 31.6 out of 35 points and achieved a perfect score of 25 out of 25 in the user evaluation.
Based on the combined results, LG AI Research, Naver Cloud, SK Telecom and Upstage initially ranked within the top four teams.
Naver Cloud ranked within the top four in combined scoring, it was judged not to meet the requirements for designation as an independent AI foundation model.
The ministry said the first-round evaluation combined AI benchmark testing, expert assessment and user evaluation, examining model performance, real-world applicability, cost efficiency relative to model size, and spillover effects on the domestic and global AI ecosystem.
Copyright ⓒ Aju Press All rights reserved.
