Cervical Cancer Identification is very critical for improved outcomes in patients and reduced mortality across the globe. Traditional screening procedures often lead to late diagnosis and very expensive treatment processes due to limitations in accuracy and efficiency. This paper introduces CoAtNet, a hybrid deep learning model that puts together convolutional neural networks with transformer-based coat nets for improved cervical cell image classification and cancer prediction. This study attempts to automate detection, which will benefit early diagnosis and treatment initiation. CoAtNet shows promise in reliably identifying key cervical cell types, which is critical for early identification and treatment. Compared to existing models utilizing the SIPAKMED dataset, CoAtNet outperforms CerviFormer, CNN-based FL, and [Hybrid CNN] ResNet50 models with an accuracy of 96.66%. These data demonstrate CoAtNet’s efficacy in automating cervical cancer diagnosis, which improves patient outcomes.