If you have a GPG private key that is locked with a password, "cracking" the key reduces to password guessing. There's a easy equation to answer that; what's hard is that there are two unknowns: the number of bits of entropy in the password and the number of guesses per second at attacker can make.
Example: There are 96 printable ASCII characters. Log2(96) is about 6.6. Given an eight character ASCII password, truly randomly generated, one would have 6.6x8=52.8 bits of entropy. If an attacker can make a billion (230) guesses per second, it will take about 222.8 seconds or about 84 days to try all combinations. Since we expect an attacker to "hit" about halfway through, average time to crack will be around 42 days.
If, instead, the password is one of the top 10,000 (and crackers have lists!) you have about 13.3 bits of entropy and cracking will succeed in less than a second.
Finally, suppose you used Diceware with a 6 word pass phrase. You'd get 66 bits of entropy and cracking would take about 235 seconds, or over a thousand years.
You can repeat the arithmetic with other values of guesses per second.
There are two important things to glean from this. First, if private key is locked with a password and the attacker has access to the locked private key, cracking is all about the password, and not about GPG or public key crypto at all. Second, every bit of entropy added to a password approximately doubles the time to crack it by brute force.
For an AES key derived from a password, the calculations are the same. However, an AES key does not have to be derived from a password. It can be randomly generated. For 128-bit AES, the brute-force attacker will take 2128-30-1 = 297 seconds. That's about 5,000,000,000,000,000,000,000 years.