The Colossal-AI team has successfully built a low-cost Chinese version of the LLaMA-2 large model, achieving excellent results in multiple evaluation rankings. They have open-sourced the complete training process, code, and weights, and provided the evaluation system framework ColossalEval. This solution can be used to construct large models in any vertical field, bringing more possibilities for AI applications.