LG U+ partners with Opt AI to boost AI token processing efficiency

LG U+ has partnered with AI model optimization specialist Opt AI to enhance the efficiency of AI services through token optimization technology. This collaboration aims to improve how AI models process data units, known as tokens, to allow for more requests to be handled using the same amount of computing resources.

The companies have reported initial results showing they can increase the amount of tokens processed on a single GPU by up to four times. By optimizing the computational structure of AI models for real-world service environments, the partnership seeks to reduce infrastructure operating costs, including GPU and power consumption, while maintaining model performance and response speeds.

LG U+ will focus on verifying and applying these technologies within its actual AI service and large-scale infrastructure environments. Opt AI will lead the research and development of model lightweighting and computational efficiency. The scope of their cooperation, which previously included optimizing small language models for mobile NPUs, is now expanding to include server GPU environments.