Konan Technology is positioning itself at the center of Korea’s push for sovereign Artificial Intelligence with a domestically developed large language model and a focus on practical deployment. The Seoul-based company, founded in 1999 and long known for text and vision software, began intensifying its generative Artificial Intelligence work in 2022 and became the first in Korea to deploy Nvidia H100 GPUs for training. In 2023 it released a 13.1 billion parameter large language model, establishing homegrown capacity to design and scale foundation models without relying on external developers.
In 2025 Konan introduced ENT-11, described as Korea’s first large language model that integrates general-purpose and reasoning capabilities in a single architecture. The company contrasts this with competitors such as Gemini that require mode switching. Konan says ENT-11 can operate both functions simultaneously and cut GPU usage by nearly half while maintaining advanced reasoning accuracy. Despite its smaller size at 32 billion parameters, more than 20 times smaller than China’s DeepSeek R1 at 671 billion, ENT-11 delivered comparable conversational and reasoning performance and outperformed DeepSeek in coding accuracy by 4.75 percentage points.
Beyond model design, Konan is leaning into secure enterprise deployment. Its retrieval-augmented generation platform, Konan RAG-X, connects internal corporate data with real-time external sources to improve accuracy while addressing data protection requirements. The product earned a Grade 1 Good Software certification and designation as a priority procurement item by Korea’s Public Procurement Service, signaling government confidence in its suitability for enterprise use. Executives emphasize a practical approach that blends precise vector search with long-standing retrieval expertise to support dynamic, up-to-date responses.
Adoption is expanding across energy and defense. Konan recently won a Korean military project to create a domain-specific large language model for intelligence-driven decision support in secure environments. The company’s package enables fine-tuning within a client’s infrastructure, allowing sensitive data to remain on site while customers adapt the model to mission needs.
Konan’s momentum aligns with national priorities under the Ministry of Science and ICT and the Ministry of SMEs and Startups to advance sovereign Artificial Intelligence across defense, public administration, and enterprise innovation. The company plans to extend its architecture to large action models for domain-specific multi-agent collaboration and is also targeting Southeast Asia through partnerships with universities and research institutes. By emphasizing efficiency, reliability, and sector-specific integration, Konan argues that compact, high-performing models can help define a new paradigm of smart sovereignty in Artificial Intelligence for Korea.
