Go to News

detail

* It has been translated using AI

LEE Garam
Input : 
2026-08-13 13:44:03
Updated : 
2026-08-13 15:05:45
Deputy Prime Minister of Bae Kyounghun and Minister of Science and Technology Information and Communication. [Yonhap News]
Deputy Prime Minister of Bae Kyounghun and Minister of Science and Technology Information and Communication. [Yonhap News]

In the global benchmark evaluation of the national artificial intelligence (AI) selection contest, start-ups received report cards ahead of large companies. Motif Technologies and Upstage were ahead, while SK Telecom and LG Group AI researchers were pushed back. In the information technology (IT) industry, there are mixed opinions that the performance of the AI model developed in Korea has risen to a level that can compete with the global open weight model, and that it is too early to decide victory or defeat as Korean proficiency is not reflected and there is a bench maxing controversy.

According to the Artificial Analytics Intelligence Index (AAII) of global AI evaluation agency Artificial Analytics on the 13th, Motif Technologies' Motif 3 scored the highest with 47 points. Upstage's "Solar Open 2" then received 37 points, SK Telecom's "Aidat X-K2" received 35 points, and LG AI Research Institute's "K-Exemployee 2.0" received 31 points. In December last year, LG AI Research Institute, which sat on the throne in the first evaluation of the Dokpa Mo Project, was pushed to the bottom, shocking the IT industry.

However, AAII grades are not directly related to the final results of the second evaluation. AAII synthesizes nine evaluations, including scientific coding, lengthy reasoning, and knowledge accuracy. It is an index that judges a wide range of agent work performance capabilities and convenience of job use. Since it is a method of converting a plurality of area benchmarks into a single score by grouping them into one, the range of score fluctuations increases depending on the benchmark composition.

In the second evaluation, the benchmark score is 40 out of a total of 100. Among them, AAII evaluation accounts for 25 points, and the National Intelligence Agency (NIA) benchmark evaluation accounts for 15 points. The rest is for experts (35 points) and users (25 points). An evaluation team of 200 citizens tries out four models and scores them. The Ministry of Science and ICT plans to drop one of the elite teams and advance the three to the next within this month.

#

Most Read News