Among the four elite teams participating in the second-stage evaluation of the government's national AI model selection project, the so-called "independent AI foundation model project (Dokpamo)," the Motif Technologies elite team was eliminated on the 18th.
Vice Minister Ryu Je-myeong of the Ministry of Science and ICT said this at a briefing titled "independent AI foundation model project second-stage evaluation results announcement" held on the morning of the 18th at the Government Complex Seoul in Jeokseon-dong, Jongno-gu, Seoul. With this, the elite teams of Upstage, SK Telecom, and LG AI Research will advance to stage three.
The second-stage evaluation covered LG AI Research's "K-Exaone 2.0," SK Telecom's "adotX K2 (A.X K2)," Upstage's "Solar Open 2," and Motif Technologies' "Motif 3 (Motif-3)," and proceeded with △benchmarks (40 points) △expert evaluation (35 points) △user evaluation (35 points).
The Ministry of Science and ICT explained, "For the elite teams advancing to the next stage, we plan to expand support for high-performance GPUs (B200), and through this, they are expected to focus on developing higher-performing independent AI models." It said it plans to increase the GPU support for the teams advancing to the next stage from the current level of 768 units to about 1,000 units.
In the first-stage evaluation on Jan. 15, LG AI Research's elite team received the highest overall score with 90.2 points. At that time, SK Telecom and Upstage also passed the first hurdle, but Naver Cloud and NC AI were eliminated. The NC AI elite team fell short in the first-stage evaluation, and the Naver Cloud elite team fell short in terms of independence. The Ministry of Science and ICT later held an additional call and brought in the Motif Technologies elite team, forming the current four-team structure.
The benchmark evaluation consisted of 25 points from the assessment by the global AI performance evaluation agency Artificial Analysis (AAII) and 15 points from the National Information Society Agency (NIA). Ahead of the announcement of the second-stage results, the evaluation by the global AI performance evaluation agency Artificial Analysis (AAII), released earlier, emerged as a variable. In the AAII evaluation published on the 13th, Motif Technologies' elite team model "Motif 3" scored 47 points, the highest among the independent AI foundation model project participants. Upstage's "Solar Open2 250B" scored 37, SK Telecom's "adotX (A.X)-K2" scored 35, and LG AI Research's "K-EXAONE (EXAONE) 2.0" followed with 31. In the benchmark evaluation, the overall average for the four elite teams was 22.5 points, and the score gap between first and fourth place was 4 points.
For the expert evaluation, a committee of AI experts was formed, and an assessment lasting about a week was conducted based on the documents submitted by the elite teams. The in-depth review focused on the teams' development strategy and technology (10 points), development performance and plans (10 points), and ripple effects and contribution plans (15 points). In particular, independence and the level of performance improvement based on it were reflected in the expert evaluation items. The overall average for the four elite teams in the expert evaluation was 28.8 points, and the score gap between first and fourth place was 2.4 points.
The user evaluation consisted of an AI expert user assessment (15 points) and a general public assessment (10 points), addressing AI usability and related factors. Forty-nine AI expert users, including heads of AI startups, participated in the AI expert user assessment. The general public assessment was conducted by a panel of 185 members of the general public. The overall average for the four elite teams in the user evaluation was 17.6 points, and the score gap between first and fourth place was 5 points.
An official at the Ministry of Science and ICT said, "As global competition over AI models intensifies and the cross-domain performance of AI models makes a noticeable leap, leading countries are moving to make AI models a national security strategic asset," adding, "The Ministry of Science and ICT also recognizes the need to develop AI models and is discussing a new support framework that goes beyond existing approaches with relevant ministries, and will soon prepare and announce specific details."