“From choosing AI models for you to managing costs”—KT unveils Token Factory

[Financial News] KT has unveiled an AI platform that brings together everything from AI model selection and usage tracking to cost management and security features in one place. Since models differ in performance and cost, the company aims to help businesses choose models suited to their needs and reduce token usage, easing the burden of AI transformation (AX).
KT announced on the 27th at Gwanghwamun West Building in Jongno District, Seoul, that it had unveiled KT Token Factory, which offers everything from AI model selection and cost management to security features, and that it would officially launch on the 30th. Token Factory provides around 500 AI models from Korea and abroad on a single platform and supports everything from sign-up and model selection to usage tracking, cost management and payment. The service will initially target corporate customers seeking to use multiple AI models while managing costs efficiently.
KT introduced Token Factory as a platform that applies to AI services the connectivity, billing and protection capabilities it has built up in the telecommunications business. The company said the platform connects multiple AI models and calculates costs based on usage, while also adding data protection and security features. The move reflects a growing trend in which choosing the right model for a task has become an AX challenge for businesses, given that AI models differ in their strengths and prices for tasks such as translation, summarization, coding and reasoning.
Token Factory’s core feature is the “Auto Model Router,” developed in-house by KT. It automatically selects a suitable AI model based on factors such as the type and difficulty of a question, the required quality, cost and the range of models a customer has authorized. It connects relatively lightweight models to simple tasks and high-performance models to complex analyses.
KT said the technology ranked first overall in RouterArena, a global benchmark that evaluates model-selection performance. The platform also uses prompt-compression technology designed for the characteristics of Korean. The company said this reduces the number of input tokens by cutting repetitive phrasing while preserving the question’s meaning and the quality of the answer.
KT said model routing and prompt compression can reduce the tokens needed to use AI by up to 25%. As increased AI use also drives up usage costs and operational burdens, the company aims to optimize model and token use for each task and lower the barrier to adoption for businesses.

KT highlighted data sovereignty and security as key features. It will provide AI inference services from data centers in Korea and has created an environment where KT’s own models can be used alongside open-weight models from Korea and abroad. It checks prompts entered by customers for personal and sensitive information and harmful language, and also checks AI-generated answers using its own guardrails.
KT plans to keep improving guardrails tailored to Korean-language environments and expand their capabilities so customers can adjust what is detected and the level of blocking according to their work environments and security requirements. The aim is to support the use of diverse AI models in corporate and industrial settings where security and data management are important.
KT said it had formed an official partnership with OpenRouter, a global AI model marketplace operator, the first of its kind in Korea, and secured around 500 models to connect to its platform. It will link OpenRouter’s global models with KT’s domestic infrastructure and sovereign AI models, and also plans to distribute overseas the AI models KT offers in Korea.
KT is also expanding partnerships with Uracle and Cast AI in AI coding and agents, and Timely in enterprise AI workspaces. It plans to work with members of the MOAI consortium, including Upstage Co., Ltd. and Daum, to build an ecosystem spanning infrastructure, models and services.
KT Executive Vice President Park Sang-won, head of the AX Business Division, said, “The AI market is rapidly shifting from a competition over who can build better models to one over who can help people use AI more easily and efficiently,” adding, “We will support customers’ AX transformation through Token Factory and develop it into a key platform that drives the growth of Korea’s AI ecosystem.”
[email protected] Yoon Hong-jip Reporter