Naver Unveils New AI Model with SLM Architecture, Cutting Costs 3x and Boosting Speed on July 2

According to Yonhapinfomax, on July 2, Naver unveiled a new AI model for its AI Tab search service, incorporating a specialized small language model (SLM) architecture. The model replaces the single large language model approach with role-specific SLM combinations, reducing infrastructure operating costs by up to 3x while improving response speed over 2x.

The model also integrates user-generated content from Naver's ecosystem, such as mom community forums and restaurant reviews, to enhance answer accuracy. Naver's Clarify RL training technique further reduced hallucination rates by up to 30 percentage points compared to Hypercloua X. The company also advanced multimodal capabilities via its Smart Lens feature, enabling the AI to process both text and images for search and reservation tasks.