Imagine a world where AI could unlock the secrets of our DNA, paving the way for breakthroughs in disease treatment and drug discovery. That's the promise of Geneverse, a collection of open-source, multimodal large language models (MLLMs) designed specifically for genomics and proteomics research. Unlike general-purpose AI, which often struggles with the complexities of biological data, Geneverse excels at interpreting genetic and protein information. It's like having a super-powered research assistant that can analyze gene functions, predict protein structures, and even identify marker genes from spatial transcriptomic data, all with impressive accuracy. This breakthrough is possible thanks to a unique training approach. Geneverse was fine-tuned using real biological data from the NCBI database, augmented with synthetic descriptions generated by GPT-3.5. This combined approach not only enhances the model's understanding of intricate biological processes but also allows it to generate descriptions that are both scientifically sound and easy to understand. Researchers have already begun using Geneverse to analyze gene embeddings, revealing hidden patterns and functional relationships. Its ability to generate comprehensive gene function summaries, accurate protein structure descriptions, and identify marker genes opens up exciting new avenues for research. While there are ongoing challenges related to computational resources and data complexity, the potential of Geneverse to transform genomic and proteomic research is undeniable. As the project evolves, incorporating even more advanced techniques and larger datasets, the future of AI-driven biological discovery looks brighter than ever.