Google DeepMind has released AlphaGenome 2, its next-generation DNA sequence model. The new model narrows the prediction scope from tissue-average levels to individual cell types and states and has been restructured into a conditional model, enabling a step-change improvement in variant prediction performance. It can capture disease-related cellular variant effects that were previously missed and can make predictions for unseen cell and disease states. The model was trained on a single-cell atlas of more than 107 million human and mouse cells. While retaining the core DNA function experimental metrics of the first generation, it adds DNA methylation prediction capability, can identify nearly twice as many variants that alter gene expression within cells, and also supports cross-modal data imputation to predict missing experimental data, helping researchers precisely pinpoint disease-related variants and uncover patterns of gene variant effects. The model is currently open to non-commercial research through the AlphaGenome API, with commercial research in a private preview stage. It will subsequently be combined with related agentic scientific tools to help accelerate research.