Symbolic regression (SR) is a machine learning technique used to uncover mathematical relationships in data by generating interpretable expressions. While several SR methods have been developed—ranging from evolutionary algorithms to neural-symbolic models—there remains a lack of standardized benchmarks that assess both prediction accuracy and interpretability. In this study, we present a comparative evaluation of four SR tools: GPlearn, PySR, Fast Function Extraction (FFX), and a neural network-based symbolic regressor. These methods are tested on three synthetic and two real-world datasets. Evaluation criteria include the coefficient of determination (R²), runtime, and expression interpretability. Results reveal that PySR consistently achieves higher accuracy, while FFX and GPlearn often produce more interpretable expressions. We also highlight challenges such as inconsistent input feature usage, runtime variability, and the need for clearer interpretability metrics. This work aims to support practitioners in selecting SR tools based on practical trade-offs and encourages further development of standardized evaluation frameworks for SR.