In recent years, multi-domain recommendation systems have rapidly evolved, using a unified model framework to transfer knowledge across domains. However, these methods assign the same task model parameters to all samples from a specific domain. This makes them difficult to capture differences between different sample groups within the same domain and commonalities between the same sample groups across different domains. To address this challenge, we propose a novel model with Feature-Adaptive dynamic network and Domain-Specific networks for multi-domain recommendation (FADS). In our proposed model, the feature-adaptive dynamic parameter network is applied to explore the commonalities among the samples with same certain features values across all specific domains and differences between the groups with different certain features values, while domain-specific networks are used to mine the uniqueness of each domain. To validate our approach, we conduct extensive experiments on two public datasets, and the results confirm the effectiveness of our proposed model.