[E-Daily Reporter Yun Junghoon ]KTCorporation(030200)’s in-house developed hyper-scale artificial intelligence (AI) model has become the first in Korea to pass a series of consecutive reliability tests, once again demonstrating its competitiveness in the B2B (business-to-business) market.

KTCorporation announced on the 8th that its self-developed ultra-large language model (LLM), “Mideum K 2.5 Pro,” has received the “Artificial Intelligence Trustworthiness Certification 2.0 (CAT·Certification of AI Trustworthiness)” from the KoreaInformation&Communication Technology Association (TTA).

CAT is a private, self-regulatory certification system that evaluates the governance and trustworthiness of AI adoption and operations across the board; it was established to systematically address ethical and technical risks that may arise during the widespread adoption of AI. Following the previously certified “Mideum K 2.0 Base,” KTCorporation has now obtained certification for “Mideum K 2.5 Pro” as well. This marks the first time in Korea that AI models from the same series have consistently passed trustworthiness verification.

The newly certified “Belief K 2.5 Pro” is a large language model with 32 billion parameters (32B). It is a closed model designed exclusively for B2B service applications, including △business automation, △customer service, and △internal operations.

Industry experts assess that this certification represents more than just a technological achievement; it marks a significant milestone for the domestic AI industry. Analysts note that by establishing a system for continuous improvement and verification—rather than stopping at a one-time certification—it has paved the way for the standardization of B2B-focused commercial AI models. Consequently, it is expected to significantly expand the range of AI options available to corporations and public institutions that prioritize security and reliability.

“Mideum K 2.5 Pro” also demonstrates overwhelming competitiveness compared to other models in its class in terms of performance. In the globally recognized AI benchmark “AAII (Artificial Analysis Intelligence Index) V3.0,” it outperformed comparable models from domestic and international competitors—such as Alibaba’s “Qwen-3-30B-A3B” from China and Naver’s “HyperCLOVA X-32B” from South Korea—to achieve top-tier performance.

In particular, in the “Tau-Square” benchmark—which evaluates an AI’s ability to connect to actual systems and perform tasks—it achieved a high task completion rate of approximately 87%. This means it has passed the test in the areas of “accuracy” and “stability,” which are considered most critical in real-world enterprise environments.

Park Jae-hyung, Head of KTCorporation’s AX Future Technology Institute, said “With ‘Belief K 2.5 Pro’ following ‘Belief K 2.0 Base’ in consecutively obtaining CAT certification, we expect competition among AI models in actual industrial settings to shift toward a ‘reliability-based’ approach in terms of performance,” he said. “KTCorporation will continue to develop reliable AI models that customers can use with confidence and take the lead in creating a responsible AI ecosystem that supports enterprise customers’ transition to AX.”