Most languages do (though implementation bugs are common—plenty of languages where 64-bit precision is required actually end up being implemented in such a way

Most languages do (though implementation bugs are common—plenty of languages where 64-bit precision is required actually end up being implemented in such a way that 80-bit precision is used on x86). That said, a lot of low-level languages don't: C doesn't guarantee this, for example. C doesn't even mandate IEEE 754, and it merely requires that "float" is a subset of the values of "double" which is itself a subset of the values of "long double".