long mask issue

[royce]

If you really think that your target plaintext is 32 digits long, you will need more information about digit placement and likelihood to have any hope of cracking it.

To answer the literal part of your question: by default, mask mode used Markov to try common combinations of characters first. There is no way to "randomize" the attempts, but if you want to simply try all combinations, you can use the --markov-disable option.

But that's not really going to make this attack useful. The length of your mask is ... extremely long. 10^32 combinations is not feasible to exhaust.

I beg to differ.

You can get randomized 32 byte numbers by using /dev/urandom.

The command would be:
cat /dev/urandom | tr -cd '0-9' | fold -w32 | ./hashcat64.bin -m <hashtype number> hash.txt <other hashcat options here>

What hash type would this be for anyway?  We are assuming that brute forcing a password that long is possible with hashcat.  I don't have the list handy, but there is a length limitation for every hash type.  Perhaps someone can link to that.

But the odds of getting the right number are WAY worse than winning the lottery.  In that respect, you are correct.

You can beg to differ all you'd like, but it does no future reader of this thread any good whatsoever.

No one should ever feed /dev/urandom to a password cracker (unless they're fuzzing it or something). Feeding /dev/urandom to hashcat would be the password-cracking equivalent of Bogosort. Not only is it monstrously inefficient, but because it is uncontrolled, some strings could be skipped and others repeated. It's demonstrably useless.

And to be more precise, what I meant was that there's no way to do what OP was asking for within hashcat. Doing it outside hashcat would also mean slowing down GPU(s) to the rate that a single CPU core can generate randomness. 

But a "how to do it faster" discussion is also pretty ridiculous when no one should ever want to do it in the first place. The OP and their grandchildren will die before it exhausts.