South Korean on-premises AI company Lumenrai, led by CEO Ahn Tae-min, is moving to target the country’s local large language model market.
Lumenrai supplies on-premises AI platforms tailored to Korean corporate environments. The company said it has proprietary technology to fine-tune and deploy retrieval-augmented generation (RAG) systems using vetted open-source LLMs, including Meta Llama, OpenAI GPT-OSS, Microsoft Phi, Qwen, Mistral, DeepSeek and EXAONE.
The company’s core capability, it said, is selecting and tuning the best model for a customer’s industry, data and security requirements, then deploying it in a fully offline environment. Lumenrai said it has co-developed a local LLM deployment platform with Japan’s Number One Solution since 2024 and, through that partnership, confirmed the product’s prospects in Japan.
Lumenrai plans to formally launch a Korea-optimized on-premises AI platform in the first half of 2026. Its flagship product will be a Korean version of “Microcosm,” an on-premises AI platform built on a 100% air-gapped architecture.
The platform runs fully offline so corporate data is not sent to external servers, the company said. Lumenrai said the system meets requirements tied to the AI Basic Act set to take effect in January 2026, the National Network Security Framework (N2SF) differentiated security at Level 3, a medical law ban on sending patient data outside, financial network separation and defense-industry security standards.
“The open-source LLM ecosystem is already mature,” Ahn said. “What the market needs now is the technology to precisely tune top-tier global open-source models for each customer and operate them reliably. We have focused on building that capability over the past two years, and we are launching it through Lumenrai, newly established in Korea.”
Ahn said Lumenrai aims to take responsibility for the full process, from choosing open-source models to automating operations, and to supply local LLMs that can be used immediately in the workplace at competitive prices with stable systems.
As more companies weigh adopting generative AI, on-premises deployments are emerging as a key strategy, particularly in industries where data security and regulatory compliance are critical. In those sectors, the trend of running open LLMs on in-house servers is accelerating.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.