KT said on the 5th that its in-house artificial intelligence (AI) model "Mideum K" ranked No. 1 among domestic small and midsize models on the global AI model performance evaluation platform Artificial Analysis Intelligence Index (AAII).
AAII is a global evaluation platform run by the AI evaluation agency "Artificial Analysis." It provides AI model information by aggregating results from multiple public benchmarks, rather than a single test, for major AI models worldwide.
Mideum K recorded generally strong scores across about 10 core evaluation items, including reasoning, domain knowledge, math/programming, and agent execution capability. KT said, "This objectively proves that Mideum K has global competitiveness as an 'agentic AI' that not only answers users' questions but also understands goals and independently uses necessary systems or tools to complete tasks."
In particular, Mideum K posted 87% on τ²-bench, one of the key agent performance benchmarks, earning recognition as a top model in the agentic AI field.
Oh Seung-pil, head of KT's Technology Innovation Division (executive vice president), said, "This AAII listing is an achievement that objectively proves KT's proprietary AI capabilities have reached global standards," adding, "Through agentic AI, we will become the best AI partner leading Korean corporations' work innovation and digital transformation."