LG has quietly entered one of the most contentious areas of technology, which is surprising for a company that most people still associate with OLED televisions and refrigerators. The 7.8-billion-parameter language model EXAONE 3.0, which was made available as open weights on Hugging Face in August 2024, did not come with the fanfare of a Meta launch or the breathless hype cycle that surrounds anything OpenAI works on. It appeared almost modestly, was made available for research and educational use on an open-source platform, and then began to surpass expectations.
Bilingual fluency—not just mediocre performance in two languages, but actual competitive strength in English along with what independent benchmarks have described as world-class proficiency in Korean—is the model’s defining feature. The detail that merits attention is that second section. Non-English languages are typically treated by the majority of large language models as afterthoughts, bolted on during fine-tuning with diminishing returns the farther you go from the Latin alphabet. EXAONE 3.0 was created from the ground up to handle Korean with the same seriousness as English. It was trained on about sixty million professional data cases in fields like mathematics, chemistry, patents, and code, and it has plans to reach a hundred million cases in other areas. It seems that LG was looking for evidence that a lab with its headquarters in Korea could create something globally relevant without compromising its home-language advantage, rather than just a bilingual model.
LG AI Research, the group’s specialized think tank founded in 2020 with Chairman Koo Kwang-mo’s support, is in charge of this endeavor. It is worthwhile to follow the timeline. The first multimodal text and image system in Korea, EXAONE 1.0, debuted in December 2021. In July 2023, version 2.0 was released, trained on 350 million photos and 45 million professional literature pieces. Then came 3.0, which brought with it a strategic shift toward open release that was similar to what businesses like Alibaba had been doing in China: employ open-source models to create developer ecosystems, draw in cloud clients, and create goodwill that eventually turns into revenue.
It’s difficult to ignore how well the model functions. Compared to its predecessor, LG claims a 56% reduction in inference time, a 35% reduction in memory usage, and a 72% reduction in operating costs. For businesses considering whether to adopt a model, numbers like that are crucial; adoption is frequently determined more quickly by speed and cost than by raw benchmark scores. Additionally, efficiency is where smaller players find their advantages in the competitive landscape that includes Alibaba’s Qwen, the UAE’s Falcon, and Meta’s Llama family.
Perhaps even more telling is what transpired after 3.0. The subsequent EXAONE 3.5 was the only Korean model to be named a “notable AI model” by Stanford’s Institute for Human-Centered Artificial Intelligence in its annual report. Then came EXAONE 4.0, a hybrid model that combined general and reasoning capabilities, and EXAONE Deep, a reasoning-focused variant that Microsoft’s AI Diffusion Report identified as Korea’s representative model, placing the nation third in the world after China and the United States. Within two weeks of its release, 4.0’s downloads exceeded half a million, surpassing the pace of any domestic Korean model.
Something intentional, almost methodical, is suggested by the progression. The parameter-count arms race that characterizes much of Western AI development is not what LG is pursuing. Rather, it appears to be developing gradually, demonstrating proficiency at every level, broadening language coverage (K-EXAONE, released in late 2025, covers six languages), and adding features like vision-language understanding with EXAONE 4.5. According to reports, the most recent model performed better on document comprehension benchmarks than GPT5-mini and Claude Sonnet 4.5. These are the kinds of real-world, enterprise-facing tasks where industrial AI truly pays off.
It’s genuinely unclear if EXAONE will develop into the foundation model that LG envisions, something associated with Korean AI in the same way that the company’s name used to refer to household appliances. There is intense competition, there are actual funding disparities between LG and American hyperscalers, and market dominance is not always the result of open-source goodwill. But in Seoul, it’s obvious that something is working. There seems to be less of a language barrier than there was two years ago, at least between Korean and English. Additionally, the business that is thinning it sells washing machines.
London Bilingualism's content on health, medicine, and weight loss is solely meant for general educational and informational purposes. This website does not offer any diagnosis, treatment recommendations, or medical advice.
We consistently compile and disseminate the most recent information, findings, and advancements from the medical, health, and weight loss sectors. When content contains opinions, commentary, or viewpoints from professionals, industry leaders, or other people, it is published exactly as it is and reflects those people's opinions rather than London Bilingualism's editorial stance.
We strongly advise all readers to consult a qualified medical professional before acting on any medical, health, dietary, or pharmaceutical information found on this website. Since every person's health situation is different, only a qualified healthcare provider who is familiar with your medical history can offer you advice that is suitable for you.
In a similar vein, any legal, regulatory, or compliance-related information found on this platform is provided solely for informational purposes and should not be used without first obtaining independent legal counsel from a licensed attorney.
You understand and agree that London Bilingualism, its editors, contributors, and affiliated parties are not responsible for any decisions made using the information on this website.
