Liner LLM Outperforms Competitor AI

Liner, a leader in AI search technology, has announced groundbreaking performance for its self-developed 'Liner Search LLM,' surpassing OpenAI's GPT-4.1 model and heralding a significant ripple in the AI search market. It has demonstrated superior capabilities compared to GPT-4.1 across all eight core components essential for AI search answer generation and is being lauded for redefining the future of search. The Liner Search LLM is a unique model that integrates eight core functions designed to deeply analyze and process user queries. This model was meticulously perfected through post-training with the vast amount of user data accumulated by Liner over more than 10 years, built upon an open-source foundation. This reflects Liner's profound technical know-how, going beyond simply adopting a model, to derive results optimized for real-world usage environments and patterns. Liner conducted a systematic internal verification process under conditions identical to real service environments, meticulously comparing and analyzing the performance (accuracy), processing speed, and cost (price per token) of 'Liner Search LLM' and GPT-4.1. In this comparison, which prioritized reproducibility and reliability, Liner achieved a surprising advantage. Notably, in four core components—'Category Classification,' 'Task Classification,' 'External Tool Execution,' and 'Intermediate Answer Generation'—Liner Search LLM demonstrated results that overwhelmingly surpassed GPT-4.1 in all three aspects: performance, speed, and cost. Furthermore, in the remaining four components—'Question Decomposition Determination,' 'Required Document Identification,' 'Intermediate Answer Generation with Sources,' and 'To-Do List Management'—it secured a competitive edge in at least two categories, firmly establishing the technical prowess of Liner Search LLM across all eight components. This overwhelming performance is the result of Liner's continuous efforts to advance its LLM training methodologies. The Liner Search LLM is specialized in flexibly solving problems and deriving the most accurate answers, and is expected to play a decisive role in effectively reducing the 'hallucination' phenomenon, a chronic problem in AI search. Cho Hyun-seok, Tech Lead at Liner, emphasized, "How data is trained and in what structure questions are processed is key to reducing AI hallucinations," revealing that Liner's unique learning strategy is the key to its success. The strengths of Liner Search LLM are also prominent in terms of economic efficiency. With processing costs per token 30-50% lower on average compared to GPT-4.1, it guarantees excellent operational efficiency and profitability even in environments with large-scale traffic. This signifies that Liner has secured not only a technological advantage but also business sustainability. Liner's latest achievement demonstrates that over 10 years of accumulated data and continuous R&D investment have culminated in differentiated competitiveness in the global AI agent market. Liner plans to continue accelerating its global market expansion by providing accurate search experiences optimized for research to users worldwide. The new horizons in AI search that Liner will open are highly anticipated.