---
title: LGU+ to Enter B2B AI Market Next Year
url: https://www.elseif.net/lgu-to-enter-b2b-ai-market-next-year
published: 2026-11-17T20:03:13+00:00
language: en
section: Models
source: https://www.aitimes.com/news/articleView.html?idxno=217408
organizations: LGU+, AI Times, AICC, Exio, FuriosaAI, Nvidia
publisher: elseif
---

# LGU+ to Enter B2B AI Market Next Year

LGU+ announced it will enter the B2B AI market next year, based on its achievements in AI token optimization technology. The company aims to target the corporate market by leveraging its own AI communication agent 'Exio' and capabilities developed through AI contact center (AICC) and internal AI transformation (AX), alongside government-led 'Everyone’s AI' model serving know-how.

The business will unfold through a partnership model with collaborators and the establishment of its own infrastructure. LGU+ has revealed specific achievements in token optimization across various layers: compute, model foundry, AI gateway, and agents.

In the compute layer, a 'Chip-Mix' strategy will be employed that does not depend on specific chips. This strategy combines GPUs from Nvidia and AMD with neural processing units (NPU) from FuriosaAI to reduce costs.

The model foundry layer is divided into 'inference optimization' and 'model lightweighting'. Inference optimization has doubled token generation speed through 'guess decoding', while processing efficiency has improved by approximately four times compared to the same infrastructure. The lightweighting has utilized technologies like 'knowledge distillation' and 'quantization', reducing on-device model size to around 800 megabytes (MB) while maintaining over 95% performance compared to cloud models.

An internal test showed that the dynamic query-based 'model routing' in the AI gateway layer reduced costs for general question handling by over 25% and coding agent costs by more than 30%.

LGU+'s Chief Technology Officer stated that the goal is to decrease the total cost of completing a task with AI, aiming to lead the market to provide corporate clients with optimal AI performance at reasonable costs starting next year.
