Despite the widespread use of various imaging modalities in clinical and experimental oncology without or with combined application of commercially available nonspecific contrast agents (CAs), development of tissue- or organ- or disease-specific CAs has been a continuing effort for pursuing ever-improved sensitivity, specificity, and applicability. This is particularly true with magnetic resonance imaging (MRI) due to its intrinsic superb spatial/temporal/contrast resolutions and adequate detectability for tiny amount of substances. In this context, research using small animal tumor models has played an indispensible role in preclinical exploration of tissue specific CAs. Emphasizing more on methodological and practical aspects, this article aims to share our cumulated experiences on how to create tumor models for evaluation and development of new tissue specific MRI CAs and how to apply such models in imaging-based research studies. With the results that are repeatedly confirmed by later clinical applications in cancer patients, some of our early preclinical studies have contributed to the designs of subsequent clinical trials on the new CAs, some studies have predicted new utilities of these CAs; and other studies have led to the discoveries of new tissue- or disease-specific CAs with novel diagnostic or even therapeutic potentials. Among commonly adopted tumor models, the chemically induced and surgically implanted nodules in the liver prove very useful to simulate primary and metastatic intrahepatic tumors, respectively in clinical patients. The methods to create tumor models have eased procedures and yielded high success rates. The specific properties of the new CAs could be outshined by intraindividual comparison to the commercial CAs as nonspecific controls. Meticulous imaging-microangiography-histology matching techniques guaranteed colocalization of the lesion on in vivo MRI and postmortem tissue specimen, hence correct imaging interpretation and longstanding conclusions. As exemplified in the real study cases, the present experimental set-up proves applicable in small animals for imaging-based oncological investigations, and may provide a platform for the currently booming molecular imaging in a multimodality environment.