Abstract: Techniques are disclosed for modeling a cell of an integrated circuit design. In one aspect of the invention, a full-space polynomial model is fit to cell information comprising measured data points associated with one or more independent variables such as voltage slew, capacitive load, supply voltage or temperature. Error values are generated indicative of error between the measured data points and the full-space polynomial model. The error values are used to partition the modeling space into domains. For at least a given one of the domains, a first polynomial model is generated based on a subset of the measured data points and at least one additional data point determined by interpolation from the measured data points in the subset. Error values are generated indicative of error between the measured data points of the subset and the first polynomial model.