What does the term "Large" mean in Large Language Models? How do parameters, training data, and computing resources contribute to model size? Does a larger model always provide better results, or should developers consider other factors when evaluating LLM performance?