Nota, a company specializing in AI model lightweighting and optimization technology, said Tuesday it held the "On-Device AI Optimization Contest with NetsPresso" at the Korea Computer Congress (KCC 2026) on Jeju Island.
The contest was designed to improve the execution efficiency of AI models in on-device environments using NetsPresso, Nota's AI optimization platform. NetsPresso is a platform that helps AI models run reliably even with limited resources by adapting them to various hardware environments.
Participants worked to improve performance so that AI models could operate faster and more efficiently in on-device environments. The "EcoPruner" team from Chung-Ang University took first place. The team proposed an optimization method to efficiently run large language models (LLMs) on small devices. It improved the AI model's per-token response latency by about 6.7 times and shortened the time to first response by about 4.0 times, while reducing memory usage by about 85%.
Kim Tae-ho, Nota's chief technology officer (CTO) and co-founder, who served as a judge, said, "This contest holds great significance in that participants directly experienced the entire AI optimization process based on NetsPresso and created meaningful performance improvements in real execution environments." He added, "Nota will continue to support developers and researchers in optimizing AI models more easily and efficiently across various hardware environments, and contribute to the expansion of the on-device AI ecosystem."
