Cupid includes the Frechet distribution. It doesn't have probplot function, but you can easily produce something similar:
% assume x1 is a vector of data
dist = Frechet(5,20,100); % Create a distribution object, preferably with parameter values
% in approximately the right range for your data
dist.EstML(x1); % Adjust parameter values to maximum likelihood estimates for this dataset
[obscdf,x] = ecdf(x1); % Compute empirical probability distribution
predcdf = dist.CDF(x); % Compute theoretical cumulative probabilities
figure;
plot(x,obscdf,'*');
hold on
plot(x,predcdf,'-');
