AI NewsWaveKT Launches 'Token Factory' Integrating 500 Domestic and Foreign AI Models on the 30th
Key points
- KT unveils Token Factory, an enterprise AI platform
- Model selection, cost management and trust features in one place
- Token costs cut by up to 25%; launch on the 30th
KT unveiled 'KT Token Factory,' an AI platform supporting enterprise AI transformation (AX), on the 27th. AI Times, Korea NGO News and The Elec reported that KT will launch the service on the 30th. Datanet reported that KT plans to grow Token Factory into a core platform for customers' AX and a new growth business.
Outlets described the service somewhat differently. According to Korea NGO News, Token Factory offers AI model selection, cost management, and AI trust and safety features on one platform to meet surging demand for AI tokens. AI Times introduced it as an enterprise service where users can select, manage and pay for more than 500 domestic and foreign AI models on a single platform. Bloter said the key point is that it connects multiple AI models through a single API and automatically picks the model suited to each task to cut costs. iNews24 named OpenAI's GPT-6 Astra, KT's Mi:dm K 3.0 Pro and Upstage's Solar Pro 4 as examples of models available on the one platform.
Several outlets reported the same cost savings of up to 25%, but described the method differently. EBN wrote that automatic recommendation of the company's own AI models and prompt compression technology cut token costs by up to 25%, while iNews24 explained that the platform reduces token use by automatically choosing a model suited to the nature of the work. In a headline on its on-site report, EBN said KT is aiming at a 215 trillion won global market with Token Factory. IT Chosun reported that Park Sang-won, executive vice president of KT's AX business division, said on the 27th that the yardstick for AI competitiveness is widening from performance to results.
A KT executive also explained the revenue model. According to Chosun Biz, Lee Jin-hyung, head of KT's AX business unit, said on the 27th that brokerage fees are not the main business model and that KT will earn revenue by running models directly on GPUs in its data centers to produce tokens. A Q&A by Digital Daily reported that the business combines token production on KT's own AI infrastructure with distribution of external AI models, and that its main source of revenue is the margin on tokens it produces itself. In a headline, the same outlet's on-site report said KT is moving beyond brokering models to become a producer. Tech M said in a headline that the service will start with 66 overseas and 7 domestic models, a different figure from the 500 or more reported by other outlets.
Among KT's recent AI news, Kyosu Shinmun reported on an agreement. Korea Polytechnic University signed a memorandum of understanding with KT on Oct. 21 on campus to cooperate on training practical AI talent and on the AICE program.
Glossary A token is the unit an AI model uses to process text, and many AI services bill by token usage. An API is a set of rules that lets programs exchange functions and data with each other.