>>> import dis
>>> f1 = lambda: 10 + 20
# yes, decompiling we see the sum was precalculated...
>>> dis.dis(f1)
  1           0 LOAD_CONST               2 (30)
              3 RETURN_VALUE        
# ...but still, the old constants were kept
>>> f1.func_code.co_consts
(10, 20, 30)

# actually, all intermediate values are kept
>>> f2 = lambda: 1 + 2 + 10 + 100
>>> f2.func_code.co_consts
(1, 2, 10, 100, 3, 13, 113)

# and even duplicated constants are generated
>>> f3 = lambda: 7 * 1
>>> f3.func_code.co_consts
(7, 1, 7)

# the same applies for other types of constants
>>> f4 = lambda: 'a' + 'b' + 'c' + 'd' + 'e'
>>> f4.func_code.co_consts
('a', 'b', 'c', 'd', 'e', 'ab', 'abc', 'abcd', 'abcde')

# this is probably related: named functions always have a None
# constant, while lambda functions do not; this None is probably
# a residue from named functions receiving an automatic "return None"
# when they don't have an explicit return
>>> def named():
        return 1

>>> anonymous = lambda: 1

>>> dis.dis(named)
  2           0 LOAD_CONST               1 (1)
              3 RETURN_VALUE        
>>> named.func_code.co_consts
(None, 1)
>>> dis.dis(anonymous)
  1           0 LOAD_CONST               0 (1)
              3 RETURN_VALUE        
>>> anonymous.func_code.co_consts
(1,)