Nam Hyun-woo has worked as a staff writer at The Korea Times since 2013, mostly covering business and politics. He currently belongs to the Business Desk where he covers topics such as emerging tech, AI, ICT and Korea's chaebol community. Prior to joining the team, he was the paper's correspondent for the presidential office of Korea during the Yoon Suk Yeol and Moon Jae-in administrations.
LG unveils 750 bil.-parameter frontier AI model K-EXAONE 2.0
A promotional image for K-EXAONE 2.0 / Courtesy of LG AI Research
By Nam Hyun-woo
LG AI Research said Friday it has released K-EXAONE 2.0, Korea's largest artificial intelligence (AI) model with 750 billion parameters, to the open-source community Hugging Face.
The model is intended for submission to the Korean government's National AI Foundation Model project, which aims to foster foundation models developed from scratch using domestic technology.
The 2.0 model has more than tripled its parameter count from the 1.0 version that LG AI Research previously submitted for the project. LG AI Research also adopted the Apache 2.0 license for K-EXAONE 2.0, allowing unrestricted commercial use.
The model scored an average of 70.1 across 24 benchmark evaluations, more than 10 percent higher than the 63.3 recorded by K-EXAONE 1.0. In particular, its average performance across three key coding and agentic coding benchmarks improved by 30 percent.
In global and Korean long-context comprehension benchmarks (OpenAI-MRCR and Ko-LongBench), the model scored 94.4 and 89.6, respectively, outperforming Zhipu's GLM-5.1.
In an agentic tool-use benchmark (Tau3-Bench Banking), the model scored 14.2, outperforming GLM-5.1 with 11.5 and Qwen3.5 with 13.4. It also delivered performance comparable to leading global models, including GLM-5.1, DeepSeek V4 Pro Max and Qwen3.5, in instruction-following benchmarks.
Benchmark comparison of K-EXAONE 2.0 with other global artificial intelligence models / Courtesy of LG AI Research
LG AI Research expanded K-EXAONE 2.0's language support to 10 languages — Korean, English, Spanish, German, Japanese, Vietnamese, French, Italian, Portuguese and Polish.
"It now supports the largest number of languages among Korean AI foundation models, laying the groundwork for Korea’s sovereign AI to expand into the global market," an LG AI Research official said.
“The significance of K-EXAONE 2.0 is not simply that it is a large-parameter model, but that Korean researchers independently completed the entire development process," said Lim Woo-hyung, co-head of LG AI Research.
"It means we have secured the capability to compete with global frontier models at the same scale, which is the core objective of the National AI Foundation Model project. The model’s current performance is only the beginning of unlocking its full potential as a global frontier-scale model. We will continue to enhance the model by improving data quality, conducting further training and reinforcement learning and refining inference technologies."