is LLaMa better than LLaMa 2

question

(need verify) due to LLaMa 2 was trained with more constrain datasets