You can certainly set the TolFun tolerance to something larger than the default, perhaps 1e-4. But that would not necessarily get you different precision, it just allows the algorithm to stop earlier.
By default, ga uses double-precision floating point numbers. You are free to use any data type you like, but then you have to write your own creation, mutation, and crossover functions. See the genetic algorithm options.
Alan Weiss
MATLAB mathematical toolbox documentation