BioT5+ is an advanced model designed to enhance biological research and drug discovery by integrating IUPAC names for molecular understanding, expanding bio-text and molecule data, and employing multi-task instruction tuning. It improves numerical data processing through advanced tokenization. BioT5+ is pre-trained and fine-tuned on diverse datasets, including 3 types of problems (classification, regression, generation), 15 tasks, and 21 benchmark datasets, demonstrating strong performance in most cases. The model excels in capturing intricate biological relationships, contributing significantly to bioinformatics and computational biology. BioT5+ addresses limitations of previous models by incorporating IUPAC names, expanding data sources, and using multi-task training. It outperforms other models in tasks like molecule property prediction, chemical reaction prediction, and molecule description generation. Ablation studies show that IUPAC names and additional data significantly improve performance. BioT5+ also performs well in protein-related tasks, including protein description generation and interaction prediction. Despite its strengths, BioT5+ has limitations in generalization across tasks and handling multi-modal data. Ethical considerations include the potential misuse of the model for generating harmful molecules. The model is supported by various funding sources and has been evaluated in multiple studies.BioT5+ is an advanced model designed to enhance biological research and drug discovery by integrating IUPAC names for molecular understanding, expanding bio-text and molecule data, and employing multi-task instruction tuning. It improves numerical data processing through advanced tokenization. BioT5+ is pre-trained and fine-tuned on diverse datasets, including 3 types of problems (classification, regression, generation), 15 tasks, and 21 benchmark datasets, demonstrating strong performance in most cases. The model excels in capturing intricate biological relationships, contributing significantly to bioinformatics and computational biology. BioT5+ addresses limitations of previous models by incorporating IUPAC names, expanding data sources, and using multi-task training. It outperforms other models in tasks like molecule property prediction, chemical reaction prediction, and molecule description generation. Ablation studies show that IUPAC names and additional data significantly improve performance. BioT5+ also performs well in protein-related tasks, including protein description generation and interaction prediction. Despite its strengths, BioT5+ has limitations in generalization across tasks and handling multi-modal data. Ethical considerations include the potential misuse of the model for generating harmful molecules. The model is supported by various funding sources and has been evaluated in multiple studies.