AXCXEPT/EZO-Qwen2.5-72B-Instruct

AXCXEPT/EZO-Qwen2.5-72B-Instruct is a 72.7 billion parameter instruction-tuned causal language model developed by AXCXEPT, based on the Qwen/Qwen2.5-72B-Instruct architecture with a 131,072 token context length. This model has undergone multiple tuning iterations to enhance overall performance, particularly excelling in Japanese language tasks. It achieved a score higher than GPT-4-Turbo on the Japanese MT Bench using GPT-4o as an evaluator, demonstrating strong multilingual capabilities despite its Japanese focus.