LG AI Unveils K-EXAONE 2.0, Korea's Largest Open-Source AI Model

LG AI Research has released K-EXAONE 2.0 on Hugging Face, its second model under Korea's sovereign AI project. It uses a hybrid attention MoE design with 750B total parameters and 37B active ones, making it Korea's largest AI foundation model. The model is open-sourced under Apache 2.0.

In tests, it scored 70.1 on average across 24 benchmarks, up 10% from the first version, with 30% better coding and agent skills. It beats GLM-5.1 in long-context understanding and tool use, and matches GLM-5.1, DeepSeek V4 Pro Max, and Qwen3.5 in instruction following.