In various sections of this book (notably Secs. 3.12 and 4.9), readers were warned about problems encountered with several models found in popular simulators, when the models are used for analog design. We give below several benchmark tests to evaluate MOSFET models in this context. These have been found very useful over the years, as a necessary (but not sufficient) set of tests that a model should pass before we can begin to trust it for analog work. Ideally, the tests should be quantitative comparisons to measured data; however, even if such data are not available, one can get very useful indications by running just the simulations indicated, since they will at least show whether the model being tested gives correct qualitative behavior. Some examples of how well popular models fare when put to these tests will be given along the way, but we will avoid giving numerical comparisons, as we do not want to limit our comments to specific models with specific parameter values. All tests are to be done on nonminimum geometries (long and wide channels) at room temperature. Models that fail most of these tests include, e.g., Spice model levels 1, 2, and 3.