LG AI Research Releases K-EXAONE 2.0 with 750B MoE Parameters
August 4, 2026
K-EXAONE 2.0 is an upcycled Mixture-of-Experts model featuring 750B total parameters and 37B activated per token. It supports 256K context lengths and expands multilingual capabilities across ten languages, specifically targeting Korean sociocultural reasoning.
HOW THIS AFFECTS YOU
●
builderYou can use a high-capacity MoE model for complex reasoning and agentic coding tasks with large context windows.
●
founderThis introduces a major open-weight competitor for multilingual foundation models in specific regional contexts.