LG Uplus said on October 2 it will work with AI model optimization company OptAI to develop token optimization technology that lowers GPU (graphics processing unit) and power consumption while improving AI processing efficiency. Tokens are the basic data units AI processes when interpreting queries and generating responses. LG Uplus will validate and deploy the technology through its AI service operations, while OptAI will develop lighter and faster models. Their earlier work applied lightweight technology to a small language model (sLM) based on LG AI Research’s Exaone, allowing it to run on an NPU (dedicated AI chip) in smartphones. The system delivered performance comparable to CPU-based methods while cutting power consumption by 78% and model size by 82%. The companies are now extending the research to server GPU environments, where their technology can process up to four times as many tokens on the same GPU. They also held a seminar on October 1 covering research trends, industry applications and implementation strategies. LG Uplus Chief Technology Officer, CTO Lee Sang-yeop said token-processing efficiency with existing resources is central to AI competitiveness and that the company will expand research through partnerships with diverse collaborators.