Elimination Game is an innovative benchmark framework for evaluating the performance of large language models (LLMs) in complex social environments. It simulates a multi-player competitive scenario similar to "Werewolf", testing the model's social reasoning, strategy selection, and deception capabilities through open discussions, private communications, and voting elimination mechanisms. This framework not only provides an important tool for researching AI intelligence in social games but also gives developers the opportunity to gain insights into the potential of models in real-world social scenarios. Its main advantages include a multi-round interactive design, dynamic alliance and betrayal mechanisms, and detailed evaluation metrics, enabling a comprehensive assessment of AI's social capabilities.