Evolution and development of opensource codes have certainly invited a much need comparison of their performances with the commercial ones. The comparisons are definitely based on the applicability, in-built tools and solution accuracy.
There has been a number of articles, discussion threads in cfd-online and elsewhere detailing these differences. I came across one such thread wherein CFD'ists evaluated the opensource codes with commercials guys like Fluent, CFX in terms of portability, industrial applicability, multi-physics capability and use in research activities...
In particular, aspects of geometry environment, CAD import, Meshing, Pre/Post processing, physical model implementation to name a few were brought into the conversation which drew very wide interest among several visitors to the popular cfd site. Inclusion of "professional" items within opensource codes, indeed, dominated the discussion. Several authors unanimously pointed to the primary differences between the opensource and commercial codes as being
1. ease of use
2. pre / post processing input and output data resp.
4. accuracy of simulations
5. simulation speed and scalability
6. additional priorities in handling industrial needs
7. availability and applicability of physical models
In my opinion, opensource codes such as OF provide, to a large extent, flexibility to the user (if the user is more towards model development and numerics) while still requiring some definite help from other opensource / commercial tools for pre/post processing (which is of more importance for an application engineers / decision makers...:) ). Certainly, in the recent times, open source codes have been well embraced by several research institutions and industries and they have started exploring the idea of fine tuning the code to suit their specific needs - a scenario not quite possible or may take long time to establish using commercial codes.
A note by Nelson of innovative-CFD on several other opensource development and application presents interesting overview of the CFD consulting world.
I came across some comparison tests between OF and other codes.
A nice presentation of the simulation of the implementation of Rotor-Stator interfaces in openFoam and comparisons with Numeca and CFX from the openFoam workshop held in 2008.
An recent journal article on the comparison of openFoam with commercial code fluent in capturing bubble columns is a good starting point for multiphase researchers.
an cfd-online thread on comparison tests including turbulent effects
Basic comparisons: flow past cylinder cases
STARCD vs OF : http://www.cfd-online.com/Forums/openfoam-solving/59753-starcd-vs-openfoam.html
The search for the "perfect" software which does not drill into our pockets is still on....