I really can't answer more precisely than that without knowing what you consider to be a "good" model, and what you are interested in modeling. Perhaps a marginally more useful answer is that they can be _excellent_ metabolic and cellular models. It's a case of "if you know what your looking for, you can pick animal models that effectively have exactly what you're looking for". The "what", here, would be a metabolic pathway, for example.
My question was what percentage of results in mice generalize to humans?
To confirm I'm interpreting your answer directly. It depends on how the scientists use the tool. It is good at verifying certain aspects but not others.
Accuracy is very high when you are studying a metabolic pathway that is (near)identical in humans and in mice.
>To confirm I'm interpreting your answer directly. It depends on how the scientists use the tool. It is good at verifying certain aspects but not others.
Yes. More precisely: animal models are accurate when the metabolic machinery in humans is also found in the animal model.