Hi S.! This is kind of a philosophical question... in a classical sense, it all depends on the spatial separation between molecules, because that determines how large the interactions between them are. For a liquid, it does not make much sense to compare molecular properties with those from the gas phase. But think of a molecule floating in space, with none other around it to interact - this certainly is a (long-lived!) case where you can very well treat it as an isolated entity.

Of course, you can think up endless examples of intermediate cases with weak interactions which may or may not be important to consider, depending on what you want to study (and how precisely). In thin gas phases like molecular beams which are often used as an experimental tool, the approximation may still be good, depending on things like the composition and density of your sample. But I fear there's really no definite answer to your question, because you will pretty much have to reconsider it specifically for every application.

On top of Matthias said; remember that while doing QM calculations we usually make even more assumptions, like that the molecule is static. This is simply not true, since time can't be stopped and each molecule is at least vibrating (and in most cases also rotating and translating). In this sense it's better to think of a molecule as a dynamic system that is constantly changing and not as a static systems with bonds and angles set to one value.

This sounds as if any calculations are doomed to fail - but this is not the case, if you are aware of all the assumptions you are making. when performing them. If you e.g. do geometry optimization of a single molecule with no solvent and find a particular bond length of 1.2387 Angstroms and will try to convince me that this is the true lenght of this bond in a low-concentration solution, I'd say you're wrong. But if you tell me that this particular bond length of the same compound in solution is likely to be 1.24 +/- 0.02 Angstroms, I'd say you are most probably right, unless the solvent forces some drastic changes in the solute. See that even though you don't have great accuracy, you can still predict many useful properties of that system based just on a single molecule calculations (geometry and vibrations).  

You always have to proceed with caution and use every scrap of chemica/physical knowledge when doing any calculations. 

Thank you Mathias and Bartosz Trzaskowski,

I am doing the quantum calculations, many asked me this question and challenged me that one cannot calculate properties of the molecule considering it as single entity. I used to admire, if is the case what is the implications of so many approximation theories in quantum meachanics ? they lead us to the closer answer which no theory proved accurately. 

"(...) challenged me that one cannot calculate properties of the molecule considering it as single entity."

Well, that's somewhat true, and you have to be carefull when translating the properties of a single molecule to a bulk scale.

But let's try an analogy. You are probably aware that when synthesizing a new organic compound and characterizing its properties, it's required to perform several analyses, including (but not limited to) NMR spectra, mass spec and crystal structure. The two first are used to identify the compound: the third as well, but also to get some information about it, like bond lengths, conformations, possible intra- and intermolecular interactions and so on. Now, in most cases we claim that the properties found in the crystal structure are similar to those in the solution - but can we really obtan properties of the molecule in solution, considering it at a completely different solid phase?

The answer is similar to the answer to the question you posted.