KT (030200.KS) said on the 28th that AutoModelRouter, its in-house AI model routing technology, ranked second overall in Router Arena, a public benchmark dedicated to large language model (LLM) routers.
Router Arena is an LLM router evaluation platform developed by researchers at Rice University in the United States, and the related research was accepted as a full paper at ICLR 2026, an international machine learning conference. Routing is a technology that automatically selects the model best suited to a user's request from among multiple artificial intelligence models.
The many AI models now available to companies have different strengths in areas such as translation, summarization, coding and reasoning, and they also differ in performance and usage costs. As a result, technology that selects the appropriate model according to the nature of a task and the required quality — rather than handling every task with a single model — is emerging as a core capability.
AutoModelRouter analyzes a user's request in terms of task type, difficulty and knowledge domain, then connects the request to a model based on a routing policy that takes into account each model's response quality and usage cost.
KT said it applied a model selection policy designed to meet the target quality companies require while also improving cost efficiency, rather than uniformly choosing either the highest-performing or the cheapest model.
For relatively simple tasks such as translation or fact-checking, the system uses a cost-efficient model that meets the required quality, while tasks requiring specialized analysis or high-level reasoning are routed to a high-performance model. Companies can reduce the burden of comparing and selecting among multiple AI models one by one, and can configure their AI operating environment more efficiently in line with the requirements and budget for each task.
AutoModelRouter is also used for the model routing function of Token Factory, a project KT is pursuing. As Token Factory integrates the operation of various AI models and token usage environments, AutoModelRouter automatically selects the model that matches a user's request, optimizing both the quality of AI services and token usage costs.
"AutoModelRouter is a technology that implements this AI orchestration capability, and it will become a core technology underpinning the competitiveness of KT's agentic AI services, including Token Factory," said Kim Jun-seok, senior vice president of KT's Agentic AI Lab. "Based on competitive AI platform technology, we will lead our corporate clients' AX, or AI transformation."
