Verified, then logged.
Sun 22 Nov 2026 · 00:34 KST / 15:34 UTC한국어 | ENSearch /

08 / Industry · Analysis

LG Uplus targets AI service costs with token efficiency

Matching AI models to task difficulty and using computing hardware efficiently underpin LG Uplus's strategy to cut operating costs while maintaining service quality.

A research monitor displays a historical time-series curve and future forecast paths linked by bidirectional arrows to blurred prose-like blocks. AI-generated image

LG Uplus, a South Korean telecommunications company, presented token-efficiency technology and related research aimed at lowering AI service operating costs, DigitalToday reported. Tokens are the units of text an AI model processes.

According to the report, the company's goal is to maintain service quality while reducing costs through model selection and efficient use of computing hardware. The approach puts both the choice of model and the hardware used to run it at the centre of service cost management.

DigitalToday reported that LG Uplus outlined the technology, research findings and plans for commercialisation at a recent briefing for journalists. Its presentation brought together research and techniques developed through service operations.

Matching models to task difficulty

According to DigitalToday, LG Uplus presented selecting an appropriate AI model for each task's difficulty as a way to reduce costs. Model selection is therefore part of the company's proposed approach to maintaining quality within a lower operating budget.

The report also described a strategy for using graphics processing units (GPUs) and neural processing units (NPUs) efficiently. These are processors used for AI computation; their use sits alongside model selection in the company's cost-reduction strategy.

Research informed by service operations

DigitalToday reported that the briefing included a paper accepted at a workshop associated with NeurIPS, an AI research conference. The paper formed part of the research LG Uplus presented on token efficiency.

According to the report, LG Uplus also shared techniques and expertise acquired through running services. Alongside its commercialisation plans, that operational experience connects the research presented at the briefing to the company's stated aim: reducing the cost of delivering AI services while maintaining their quality.