LG AI Research Launches Exaone 4.0 for Enterprise AI Solutions

On 15 July, LG AI Research—the AI R&D arm of South Korea’s LG Group—unveiled Exaone 4.0, a hybrid reasoning AI model combining general language processing with advanced reasoning capabilities. The model outperforms competitors like Alibaba, Microsoft, and Mistral AI in science, math, and coding benchmarks but falls short of Deepseek’s top model.

Targeting Business Users

Unlike consumer-focused models like ChatGPT and Gemini, Exaone 4.0 is designed for enterprise use. "Our primary focus is on the business-to-business (B2B) sector," says Honglak Lee, co-head of LG AI Research. The model is available for research and academic use on Hugging Face and now supports Spanish alongside Korean and English.

Expanding the Exaone Ecosystem

At the AI Talk 2025 event on 22 July, LG AI Research revealed its strategic roadmap, including:

• Exaone 4.0 Vision Language: A multimodal model for text and image interpretation.
• Exaone Path 2.0: A healthcare-focused model for rapid patient diagnosis.
• Enterprise AI Agents: Including ChatExaone for internal workflows and Exaone Data Foundry for accelerated data generation.

The company claims Exaone 4.0 VL outperforms Meta’s Llama 4 Scout, and Data Foundry can achieve in one day what typically takes 60 experts three months.

Hardware and Energy Efficiency

Exaone’s on-premise agent runs on NPUs from FuriosaAI, delivering 2.25x faster inference than GPUs. The RNGD accelerator also boasts 3.75x more tokens per rack within the same power limits.

Future Goals

LG AI Research aims to provide end-to-end systems for autonomous enterprise agents, with plans for physical AI integration in robots. "We’re building web agents that can navigate complex information flows and make autonomous decisions," says Lee.

Tags: #AI #Enterprise #LG