The global benchmark report card is out for corporations participating in the second-round evaluation of the government's national-team AI model competition, the "independent AI foundation model project" ("Dokpamo"). Motif Technologies' "Motif 3" led among Korean corporations with 47 points, followed by Upstage, SK Telecom, and LG AI Research. Korean corporations' rankings remain low compared with the performance of the United States and China.
On the 13th, according to global AI performance evaluator "Artificial Analysis," Motif Technologies' "Motif 3" scored 47 in the latest Artificial Analysis Intelligence Index (AAII), the highest among Dokpamo participants. Upstage's "Solar Open2 250B" scored 37, SK Telecom's "adot X (A.X)-K2" scored 35, and LG AI Research's "K-EXAONE (EXAONE) 2.0" followed with 31.
Im Jeong-hwan, CEO of Motif Technologies, said, "In the AI model market, borders are meaningless, and only delivering world-class performance is meaningful as an independent foundation model," and added, "Based on the confidence we gained this time, we will complete a top-tier AI model that can compete on equal footing with frontier models from the United States and China."
Artificial Analysis is a global AI analytics organization that comprehensively evaluates capabilities such as math, science, coding, and reasoning in AI models.
Compared with top overseas models, domestic models do not rank high. Anthropic's "Claude Opus 5" at the highest reasoning setting leads overall with 63 points. OpenAI's "GPT-5.6 Sol" and SpaceXAI's "Grok 4.6" each scored 61. Chinese AI companies' pursuit of U.S. models stands out. Moonshot AI's "Kimi K3" has 60 points, trailing GPT-5.6 Sol by 1 point. Alibaba's "Qwen3.8 Max" also scored 58, ranking sixth overall. In addition, Z.ai's GLM-5.2 has 53 points, and the latest DeepSeek V4 Flash model also has 52.
However, the "Artificial Analysis" results do not carry over to the second Dokpamo results. In addition, Korean-language performance is not reflected in Artificial Analysis, which is a limitation.
In the second Dokpamo evaluation, benchmark scores account for 40 out of 100 points. Of that, the Artificial Analysis Intelligence Index (AAII) accounts for 25 points, and the National Information Society Agency (NIA) benchmark evaluation accounts for 15 points. The remaining points are composed of expert evaluation (35 points) and user evaluation (25 points). On Aug. 8–11, a public evaluation panel of 200 citizens used and evaluated the four models.
The government will soon release the second Dokpamo evaluation results. In the first evaluation, Naver Cloud and NC AI were eliminated, and in the second evaluation, one of the four teams—LG AI Research, Upstage, SK Telecom, and Motif Technologies—will be dropped.