Most existing facial age estimation methods operate as isolated single-task systems, neglecting the potential synergies among correlated facial attributes. We propose Dynamic Routing Age (DRAge), a novel facial age estimation approach that integrates dynamic routing and multi-task collaboration. Our framework employs the Dynamic Routing Mixture of Experts (DRMoE) architecture, where features are dynamically routed to task-specific experts via cosine similarity between input representations and expert embeddings, ensuring adaptive feature selection. Additionally, we incorporate multiple task-specific subnets. We design a Cross-Attention Mechanism Feature Fusion (CAMFF) module within the facial analysis subnet, which strengthens feature representation through attention-driven interactions. To leverage pretrained features from related tasks while avoiding feature interference during multi-task optimization, we freeze other face analysis subnets during training and fine-tune specifically for age estimation. The effectiveness and feasibility of the proposed model was validated through extensive age estimation experiments on the FG-NET and CLAP2015 facial image datasets.