It does, you can read allele frequencies using the --read-freq flag. This is relevant when computing pair-wise genome-wide IBD on a large dataset and splitting work on mutiple processes using the --genome-lists flag. Full details on the documentation:

http://pngu.mgh.harvard.edu/~purcell/plink/strat.shtml#cluster

Are you running something like --genome? If you are running on the ped/map type files PLINK should detect the allele frequencies based on your data. If your data isn't representative then the allele frequencies will be skewed anyway.