The default value for the 'maxBins' parameter in Spark MLlib's Decision Tree implementation is 32. This setting strikes a balance between accuracy and computational efficiency, making it a practical starting point for most datasets. It discretizes continuous features into 32 bins, which is sufficient for many scenarios without unnecessarily increasing computation time. Other values like 16, 64, or 128 can be used based on specific requirements, but 32 is the standard default. The option suggesting a variable amount based on categorical columns is incorrect because 'maxBins' is specifically for continuous features, not categorical ones.