It really depends on what your qustion is. Zebrafish would be a great way to add or remove genes and look at effects on gut organisation, so may be good for genetics studies, however >70% of colon cancer risk is environmental. Needless to say the zebrafish environment is very different from human and would be a poor model of sdet or physical activity for example.
I emit serious doubts about the fact that zebrafish "colon cancers" would respond to chemotherapeutics used in humans ... It is already very hard to predict this feature from genetically engineered rodents.
As a "first" model syngeneic colon cancer in immunocompetent mice would be a rationale approach. The mouse CT26 model grows s.c. as well as orthotopically (caecum).
The human HT29 orthotopic xenograft model (implanted in the caecum of immunocompromized (nude) mice) is also a "good" model.
A model is a model and remains a model. It will never translate the actual clinical situation. You have to make a choice for a model that can help you to scientifically demonstrate your working hypothesis.