SKT 'A.X-K1', 수학 추론서 글로벌 동급 모델 앞섰다

SK Telecom's A.X-K1 achieves the highest math reasoning performance among Korean independent models, outperforming certain global peers in pass@1 accuracy (83.75%), signaling that Korean AI foundation models are closing the gap with global leaders. However, the significant gap with the latest global models reveals that the focus must shift from pure model scaling to designing systems that integrate verification, exploration, and proof mechanisms. This marks a critical inflection point where AI competition moves from raw capability to systematic reasoning architecture.