This study evaluates the efficiency of code generation by Large Language Models (LLMs) and measures their performance against human-crafted solutions using a dataset from Leetcode. We compare 18 LLMs, considering factors such as model temperature and success rate, and their impact on code performance. This research introduces a novel method for measuring and comparing the speed of LLM-generated code, revealing that LLMs produce code with comparable performance, irrespective of the adopted LLM. We also find that LLMs are capable of generating code that is, on average, more efficient than the code written by humans. The paper further discusses the use of Leetcode as a benchmarking dataset, the limitations imposed by potential data contamination, and the platform's measurement reliability. We believe that our findings contribute to a better understanding of LLM capabilities in code generation and set the stage for future optimizations in the field. The study investigates the performance of code generated by 18 LLMs on 204 problems and investigates performance differences across models using a novel method for measuring and comparing the performance of LLM-generated code. We compare the performance of the code generated by LLMs to the code written by humans. Incidentally, we evaluate the usability of Leetcode, a public repository of algorithmic problems that we use as a dataset. The methodology involves generating code for each problem using different LLMs and temperatures, validating the generated code for correctness and performance, and measuring the run time of the generated solutions. The results show that LLMs can generate code that is, on average, more efficient than human-written code. However, the performance of LLMs varies depending on the temperature setting, with higher temperatures leading to greater variability in performance but not necessarily better or worse solutions on average. The study also highlights the challenges of using Leetcode as a benchmarking dataset, including data contamination and measurement reliability. The results suggest that while Leetcode's problems are suitable for performance evaluation, their measures should be used cautiously due to issues with stability and reliability. Additionally, the use of newer Leetcode problems is essential to avoid data contamination and ensure the validity of LLM evaluations. The findings of this study contribute to a better understanding of LLM capabilities in code generation and set the stage for future optimizations in the field. The study also highlights the importance of considering factors such as model temperature and success rate when evaluating the performance of LLMs. The results suggest that while LLMs can generate code that is, on average, more efficient than human-written code, there are still challenges in ensuring the reliability and accuracy of the generated code.This study evaluates the efficiency of code generation by Large Language Models (LLMs) and measures their performance against human-crafted solutions using a dataset from Leetcode. We compare 18 LLMs, considering factors such as model temperature and success rate, and their impact on code performance. This research introduces a novel method for measuring and comparing the speed of LLM-generated code, revealing that LLMs produce code with comparable performance, irrespective of the adopted LLM. We also find that LLMs are capable of generating code that is, on average, more efficient than the code written by humans. The paper further discusses the use of Leetcode as a benchmarking dataset, the limitations imposed by potential data contamination, and the platform's measurement reliability. We believe that our findings contribute to a better understanding of LLM capabilities in code generation and set the stage for future optimizations in the field. The study investigates the performance of code generated by 18 LLMs on 204 problems and investigates performance differences across models using a novel method for measuring and comparing the performance of LLM-generated code. We compare the performance of the code generated by LLMs to the code written by humans. Incidentally, we evaluate the usability of Leetcode, a public repository of algorithmic problems that we use as a dataset. The methodology involves generating code for each problem using different LLMs and temperatures, validating the generated code for correctness and performance, and measuring the run time of the generated solutions. The results show that LLMs can generate code that is, on average, more efficient than human-written code. However, the performance of LLMs varies depending on the temperature setting, with higher temperatures leading to greater variability in performance but not necessarily better or worse solutions on average. The study also highlights the challenges of using Leetcode as a benchmarking dataset, including data contamination and measurement reliability. The results suggest that while Leetcode's problems are suitable for performance evaluation, their measures should be used cautiously due to issues with stability and reliability. Additionally, the use of newer Leetcode problems is essential to avoid data contamination and ensure the validity of LLM evaluations. The findings of this study contribute to a better understanding of LLM capabilities in code generation and set the stage for future optimizations in the field. The study also highlights the importance of considering factors such as model temperature and success rate when evaluating the performance of LLMs. The results suggest that while LLMs can generate code that is, on average, more efficient than human-written code, there are still challenges in ensuring the reliability and accuracy of the generated code.