The native structures are slowly being released for CASP10 targets; all of them will be available by the end of November. In the meantime, you can look at the results of a much larger scale test of prediction methods called CAMEO. CAMEO takes newly solved protein structure before they are published, and sends the amino acid sequences out to structure prediction servers. This happens every week, so it is great to assess new methods as they are being developed. You can look at the results, as well as get more information about CAMEO, at

http://beta.cameo3d.org/modeling/weekly_summary.html

The best number to compare is the "Average accuracy (all targets)" as some servers only model the easy ones. The good thing about CAMEO is that there are many more test cases than CASP, and also that results are released each week so we can see what is working well and what needs to be improved. You will see that ROBETTA, which is now using some of your computing resources, is doing pretty well recently; before this we had problems with some targets not getting enough work units before the server deadline.