LG Uplus and OptAI team up on AI token optimization

LG Uplus and Korean startup OptAI will jointly research token optimization to improve AI service efficiency. On GPU they raised processed token volume up to 4x, while on-device EXAONE-based small language models cut power use by 78% and model size by 82%.

• GPU token throughput increased up to 4x on the same hardware
• On-device EXAONE sLM cuts power by 78% and size by 82%
• Partners held an Efficient AI seminar with FuriosaAI, Vaiseul AI, KETI and Hanyang
• Goal is to lower operating costs of large-scale AI services