본문바로가기

Home LLM · RAG-X Konan LLM

Why you should opt
Konan LLM

Konan LLM ENT-11, Korea's first unified model
for generation and reasoning,
outperforms DeepSeek-R1 while running stably
even on lower GPU specs.

Konan LLM is a Large Language Model independently developed by Konan Technology.
Excelling over DeepSeek-R1 in eight categories such as writing, roleplay, reasoning,
and coding, Konan LLM ENT-11 unifies generative and reasoning modes for more
refined and flexible performance.

Korea’s Leading Foundation Model, Konan LLM:
From Start to Present


December 2022 Initiated in-house development of a large language model; placed Korea’s first NVIDIA H100 order
June 2023  Received Korea’s first NVIDIA H100; accelerated development with timely GPU acquisition
July 2023  Completed training of a 13 billion parameter model
August 2023  Held launch press conference; introduced as a Korean on-premise LLM
September 2023  Demonstrated product at Korea’s AI Declaration Ceremony; conducted PoC with Ministry of the Interior and Safety
December 2023  Continued model advancement; completed training of a 41 billion parameter model
June 2024  Signed ₩4.1 billion contract with Korea Southern Power; first public sector deployment of a domestic LLM
July 2024  Listed on Stanford University’s ecosystem graph as a high-quality Korean foundation model
September 2024  Product lineup refined into OND (On-Device), PRO (Professional), and ENT (Enterprise)
February 2025  Co-developed Korea’s first LLM-based medical platform with Hallym University Medical Center
March 2025  Released ENT-11, Korea’s first unified model for generation and reasoning; superior to DeepSeek in reasoning tasks
May 2025  Featured case studies at the 2025 Konan Technology AI Showcase; unveiled first AI PC with built-in Konan LLM

성능 평가 결과
  • Konan LLM 32B (ENT-11)
  • DeepSeek-R1-32B
  • DeepSeek-R1-671B

자체 성능 평가 점수



BENEFITS

Benefit 01.

Advanced Reasoning Beyond DeepSeek

Designed to understand context precisely even in complex or low-data situations, producing high-quality, logic-based responses.


Benefit 02.

Zero Risk of Data Leakage

ENT-11 operates reliably on-premise or in closed networks, making it safe for use in highly regulated industries.


Benefit 03.

Structured Reasoning for Greater Trust

Built to reveal its reasoning process, allowing users to verify answers and trust the AI.


Benefit 04.

Stronger Security Through Granular Access Control

Easily control and secure sensitive information based on user roles.

FEATURE

Three Unique Advantages
of Konan LLM

RAG-X-Based LLM that Finds Answers from All Available Information

  • Integrated with Konan’s proprietary vector search technology, the LLM understands the context of the latest documents and generates responses based on accurate references.
  • The Retrieval-Augmented Generation (RAG) architecture minimizes hallucinations while enabling controlled access to information depending on user permissions.
  • It connects to internal documents and databases in real time, delivering both accurate and practical responses tailored to enterprise data assets.

Flexible LLM Lineup Optimized for Every Business Environment

  • Konan LLM offers OND, PRO, and ENT models, allowing businesses to choose the most suitable option based on size, goals, and budget.
  • From lightweight PC deployments to large-scale GPU clusters and servers, Konan LLM supports flexible, customized operation based on performance needs.

Trained on Korea’s Largest Korean Language Dataset – Exceptional Korean Performance

  • With over 25 billion data points collected through the PulseK AI analytics service since 2007, Konan LLM is trained on high-quality Korean datasets.
  • It demonstrates superior performance in context comprehension, instruction following, and consistent writing style in Korean.

Why Konan LLM
Excels in Document Tasks

 An LLM that Thinks and Explains Like a Human

Trained on high-quality data such as legal documents, patents, reports,
and news, handles complex queries with step-by-step reasoning and explanatory answers.

답변 생성 결과 비교
(코난 LLM ENT-11 vs DeepSeek 32B)

 Tailored Models to Fit Each Customer

Clients can choose the most suitable model based on their size, budget, and application needs.

 Safe from Data Breaches

As an on-premise solution, Konan LLM ensures data safety and
uses personal information filtering to prevent leakage.

 Answers Based on the Most Reliable Sources

It delivers accurate, timely responses
by referencing both internal corporate data and external up-to-date sources.

 Document Access Control by User Role

By verifying access rights per user, it reduces leakage risk and enables
document generation for authorized users.

코난 LLM 파운데이션 모델의 서비스 과정

 Easy Model Management and Evaluation

Using the no-code MLOps tool 'Konan LLM Studio',
administrators can manage models and conduct performance evaluations with ease.

코난 LLM 스튜디오 기능



USE CASES

Konan LLMsupports digital transformation wherever your data lives.
Organizations such as Korea Southern Power and Hallym University Medical Center
have already adopted it for AI-driven innovation.
Successful PoCs in both public and private sectors are leading to full-scale implementations.
Custom training and fine-tuning for each business result in meaningful productivity gains.



[대표 구축사례]




공공기관
Government Services


금융 / 보험/ 의료
Finance / Insurance / Medical care


항공 / 제조
Aviation / Manufacturing

BUSINESS

Upgrade Your Business with Konan LLM:
Balancing Performance, Cost, and Security!

Terms of Service for Konan LLM

Talk to KONAN

Do you have questions about the product?

Contact