Objective To determine the application value of the mono-exponential model, dual-exponential model, and stretched-exponential model of MRI with diffusion-weighted imaging (DWI) in breast cancer (BC) lesions. Methods Totally 64 cases with BC admitted to our hospital between June 2019 and October 2020 were enrolled in this study. They had 71 lesions in total, including 40 benign tumor lesions (including 9 breast cyst lesions) and 31 malignant tumor lesions. After DWI examination, with normal glands as control, mono-exponential model (ADC) map, dual-exponential model (Standard-ADC) map, slow apparent diffusion coefficient (Slow-ADC) map, fast-apparent diffusion coefficient (Fast-ADC) map, and stretched-exponential model (DDC) map were processed, and corresponding values were generated. Then, the situation and significance of each parameter in breast cysts, benign breast tumor lesions, and malignant tumor lesions were analyzed. Results The values of ADC, Standard-ADC, and DDC of breast cysts were higher than those of normal glands (all P < 0.05), and the values of ADC and DDC of benign breast tumor lesions were lower than those of normal glands ( P < 0.05). In addition, malignant breast tumor lesions had lower values of ADC, Standard-ADC, Slow-ADC, and DDC and a higher Fast-ADC value compared to normal glands (all P < 0.05). Compared with benign tumor lesions, malignant tumor lesions had lower values of ADC, Standard-ADC, Slow-ADC, and DDC and a higher value of Fast-ADC (all P < 0.05). Moreover, the receiver operating characteristic (ROC) curve-based analysis revealed that all the above models could be adopted to effectively evaluate the deterioration of benign breast tumor lesions (all P < 0.05), and DDC value had the most significant diagnostic effect on malignant tumor lesions ( P < 0.05). Conclusion Both dual-exponential model and stretched-exponential model of DWI can help effectively evaluate the progression of benign breast tumors, and the stretched-exponential model is more effective in the diagnosis of malignant breast tumors. These models are of great help to the future clinical diagnosis of BC.