Exaone ecosystem spreads with new AI model

On 15 July, LG AI Research,AI R&D Arm South Korea’s LG Group,Unveiled EXAONE 4.0A hybrid Reasoning AI model that combines general language processing with advanced logic capabilities initiated through the company’s first Deep model,

LG AI Research says that new model Goes ahead of the same model from Alibaba, MicrosoftAnd Mistral AI Industry for science, mathematics and coding in benchmark. However, EXAONE 4.0 is still low The best model of lampsac,

However, LG AI Research is not pursuing the same user as most familiar names in AI. Unlike models such as Chatgpt and Gemini, which are mainly designed for the average person, LG AI is targeting business users. “Our primary focus is on business-to-business (for now) on business-to-business (B2B) area,” Hanglak LeeLG AI Research in Google Brain and newly appointed co-head of former research scientist. LG launched the company in December 2020 as part of the Korean tech giant Digital Transformation Strategy.

By that end, LG AI Research has provided Exaone 4.0 for research and educational use. On a hugGlobal Open-SOS AI platform. The model now also supports the use of Spanish language, expanding its abilities beyond its original competencies with Korean and English.

Exon Ecosystem and Strategic Roadmap

Just a week after the introduction of EXAONE 4.0, LG AI Research made its commitment to the B2B focus by unveiling its broad EXAONE ecosystem and strategic roadmap. On July 22, in AI Talk 2025, the company revealed several new models.

The models are Exaone 4.0 Vision Language, a multimodal AI model that can explain both lessons and images, and Exaone Path 2.0, A healthcare-centric model Designed to diagnose the patient’s conditions in minutes. There are also many enterprise-specific AI agents: Chatexaone, An agent is currently being used internally by LG employees to support corporate workflows; EXAONE data foundry, a platform to intensify data generation; And an on-radius, full-stack agent that can be deployed in a separate, safe environment without highlighting sensitive data.

LG AI Research says that the Axon 4.0 VL, which will be launched in the near future, removes the Meta’s Lama 4 scouts in the performance tests. Additionally, the company says that data foundry can do in a single day which usually takes 60 experts three months.

EXAONE’s on-primesies agent runs on chips developed by South Korea-based Startup Manufacturing Furyosai Nerve processing units (NPU) was sewn for AI workload. According to the company, FurioSai‘S RNGD accelerator GPU’s competition was demonstrated on the exaone model 2.25 times as a competition.

LG also states that hardware is more energy-skilled. A single rack powered by RNGD chips can generate up to 3.75 times as several tokens for the exaone model compared to a traditional GPU rack within the same power range.

Autonomous agent for enterprise safety

The final goal of LG AI Research is to equip the enterprises with all the main components required to safely run the autonomous agents within their own infrastructure, with the underlying data generation and business operations facilities, Lee told the IEEE spectrum.

“We are not just offering the engine,” Lee says. “We aim to provide an end-to-end system that actually integrates the major functionality enterprises-so they can immediately plug it into their workflow. Each enterprise has unique operating requirements. So we are designed to flexible our solutions to combine and configure different parts based on each customer’s environment.”

In addition, the company is laying the foundation for the inclusion of physical AI, or AI in the robot. “Physical AI is still in its early stage,” Li says. “But the main outline in a continuous loop – the assumption, logic and action – something that we are actively manufacturing.”

While the company is not yet applying it directly to robots, they are performing the same loop with chetaxone, or Nexus agent, an AI agent, which is designed to assess the legal compliance of data sets. Nexus has the ability to crawl important internet. “These agents need to understand web pages, extract relevant insights and work on them,” Li says. “This is why we are creating web agents that can navigate complex information flow and take autonomous decisions.”