모티프, 글로벌 AI 평가서 1위?…‘AAII 점수’ 어디까지 믿어야 하나

The second round of evaluation for South Korea's 'Domestic Artificial Intelligence (AI) Foundation Model' (DomAI) project has sparked discussion as Motip Technology secured the top global AI performance score. Motip outperformed global giants like LG AI Research and SK Telecom using a smaller model size, a development that has caught industry attention.

Intelligence Analytics, a global AI performance assessment organization, reported that Motip's 'Motip 3' model achieved a score of 47, the highest among the four participating startups. The second place went to Seoul's 'Solana Open 2 250B' with 37 points, while SK Telecom's 'A.X K2' and LG AI Research's 'K-Exascale 2.0' scored 35 and 31 respectively.

Interestingly, the model size does not necessarily correlate with the evaluation score. The K-Exascale 2.0 model, with 7500 billion parameters, ranked fourth despite having more parameters than Motip 3's 3140 billion. Experts caution that relying solely on the Intelligence Index score to compare model performance may not be accurate, as it measures only a subset of a model's capabilities.