This paper introduces PythonSaga, a new benchmark for evaluating code generation models, addressing the limitations of existing benchmarks in terms of diversity in programming concepts and difficulty levels. Existing benchmarks like HumanEval and MBPP show a significant bias towards a limited set of programming concepts and predominantly focus on easy tasks, which may overestimate the performance of code generation models. PythonSaga features 185 hand-crafted prompts, evenly distributed across 38 programming concepts at three difficulty levels. The benchmark is designed to provide a more balanced and comprehensive evaluation of code generation models. The paper also presents a detailed analysis of the performance of various open and closed-source code generation models on PythonSaga, revealing that most models struggle with advanced programming concepts and often produce syntactically incorrect or incomplete code. The study highlights the need for improved benchmarks and evaluation methods that accurately reflect the capabilities and limitations of code generation models. The findings suggest that existing benchmarks may overestimate the performance of code generation models on code generation tasks. PythonSaga provides a more robust assessment framework for evaluating code generation models, paving the way for future research in this area.This paper introduces PythonSaga, a new benchmark for evaluating code generation models, addressing the limitations of existing benchmarks in terms of diversity in programming concepts and difficulty levels. Existing benchmarks like HumanEval and MBPP show a significant bias towards a limited set of programming concepts and predominantly focus on easy tasks, which may overestimate the performance of code generation models. PythonSaga features 185 hand-crafted prompts, evenly distributed across 38 programming concepts at three difficulty levels. The benchmark is designed to provide a more balanced and comprehensive evaluation of code generation models. The paper also presents a detailed analysis of the performance of various open and closed-source code generation models on PythonSaga, revealing that most models struggle with advanced programming concepts and often produce syntactically incorrect or incomplete code. The study highlights the need for improved benchmarks and evaluation methods that accurately reflect the capabilities and limitations of code generation models. The findings suggest that existing benchmarks may overestimate the performance of code generation models on code generation tasks. PythonSaga provides a more robust assessment framework for evaluating code generation models, paving the way for future research in this area.