AntiBP3 is a method for predicting antibacterial peptides (ABPs) effective against gram-positive, gram-negative, and gram-variable bacteria. The study introduces a novel approach that combines machine learning (ML) and deep learning (DL) techniques to improve prediction accuracy. The method uses alignment-free features such as amino acid composition, binary profiles, and fastText embeddings to identify ABPs. The models were evaluated using five-fold cross-validation on independent datasets, achieving high AUC values (0.93, 0.98, and 0.94 for gram-positive, gram-negative, and gram-variable bacteria, respectively). The method outperforms existing approaches and provides a user-friendly web server, standalone package, and pip package for peptide-based therapeutics. The study highlights the importance of amino acid composition and terminal residues in ABP prediction. It also demonstrates that hybrid models combining BLAST and ML or motif-based approaches do not significantly improve performance over individual ML models. The results show that ML-based models are more effective than DL models for this task. The study also compares AntiBP3 with other prediction tools, demonstrating its superior performance in terms of sensitivity, specificity, and AUC. The method is implemented as an online server, allowing users to predict ABPs for different bacterial groups based on sequence information. The study concludes that ABPs are important for combating drug-resistant bacteria and that the proposed method provides a reliable and efficient tool for ABP prediction.AntiBP3 is a method for predicting antibacterial peptides (ABPs) effective against gram-positive, gram-negative, and gram-variable bacteria. The study introduces a novel approach that combines machine learning (ML) and deep learning (DL) techniques to improve prediction accuracy. The method uses alignment-free features such as amino acid composition, binary profiles, and fastText embeddings to identify ABPs. The models were evaluated using five-fold cross-validation on independent datasets, achieving high AUC values (0.93, 0.98, and 0.94 for gram-positive, gram-negative, and gram-variable bacteria, respectively). The method outperforms existing approaches and provides a user-friendly web server, standalone package, and pip package for peptide-based therapeutics. The study highlights the importance of amino acid composition and terminal residues in ABP prediction. It also demonstrates that hybrid models combining BLAST and ML or motif-based approaches do not significantly improve performance over individual ML models. The results show that ML-based models are more effective than DL models for this task. The study also compares AntiBP3 with other prediction tools, demonstrating its superior performance in terms of sensitivity, specificity, and AUC. The method is implemented as an online server, allowing users to predict ABPs for different bacterial groups based on sequence information. The study concludes that ABPs are important for combating drug-resistant bacteria and that the proposed method provides a reliable and efficient tool for ABP prediction.