allganize/Llama-3-Alpha-Ko-8B-Instruct

allganize/Llama-3-Alpha-Ko-8B-Instruct is an 8 billion parameter language model developed by Allganize, built upon the Llama-3-8B architecture using an Evolutionary Model Merging technique. This model demonstrates strong bilingual capabilities in Korean and English, excelling in complex language tasks and logical reasoning. It achieves an impressive 6.62 on the LogicKor benchmark, rivaling the performance of larger 70B models, making it highly efficient for demanding language applications.