SK Telecom leads evaluation of sovereign AI model project in South Korea
South Korea’s Ministry of Science and ICT has released detailed results of the second-stage evaluation of a project to develop sovereign artificial intelligence foundation models. SK Telecom received the highest total score — 70.6 out of 100 — reports The Korea Herald.
Upstage ranked second with 69.9 points, while LG AI Research Institute came third with 69 points. Motif Technologies received 65.8 points and was eliminated after the second stage. The detailed results were disclosed at the request of Motif Technologies, with the consent of the other shortlisted teams.
How the score was formed
The overall result consisted of 100 points. Benchmark evaluations accounted for 40 points: 25 points were assigned to the Artificial Analysis Intelligence Index and another 15 to the National Information Society Agency benchmark. Expert evaluation carried a weight of 35 points, while user evaluation accounted for 25 points, including 15 points from professional users and 10 from the public.
More current news is available on the UA.News Telegram channel Telegram.
In the Artificial Analysis Intelligence Index evaluation, which uses global benchmarks, Motif Technologies posted the best result with 11.9 points. Upstage received 9.4 points, SK Telecom 8.8 points, and LG AI Research Institute 7.8 points. In the National Information Society Agency benchmark, which covered seven areas including Korean-language capabilities and reliability, SK Telecom scored 13.4 points and Upstage 13.3 points. LG AI Research Institute and Motif Technologies scored 12.8 and 12.7 points, respectively.
Expert and user assessments
Ten external experts from business, academia and research institutions gave the highest score to LG AI Research Institute — 29.5 points. SK Telecom received 29.3 points, Upstage 29.1 points, and Motif Technologies 27.1 points. Among 49 AI startup executives who assessed the models’ practical applicability, SK Telecom achieved the highest result with 11.6 points. In the public evaluation involving 185 people, LG AI Research Institute led with 7.6 points.
The ministry said that the project’s goal is not only to rank teams or eliminate participants, but to qualitatively develop and expand South Korea’s AI ecosystem. Development goals are reviewed every six months, taking into account the rapid advancement of technology, while the evaluation criteria and methodology were agreed upon after consultations with the shortlisted teams. Motif Technologies had previously filed an objection to the second-stage results. The ministry said it would review it in accordance with the established procedure and make a final decision within 15 days.
