From SAS:


The following approach will allow you to output a data table with all the 
models and R-squares and then subset it to the top models for each number of 
predictors. Once you have performed the ALL POSSIBLE MODELS, if you RIGHT CLICK 
in the generated results of all possible models, you should have option 
to MAKE INTO A DATA TABLE. The next step would be to add a COLUMN to this new 
data table, let's call the new column BEST RSQUARE; data type : numeric; 
modeling: continous; New Property: FORMULA; 

copy this formula into formula box:  
 
If(Row() == 1, 1, If( :Number[Row()] !=  :Number[Row() - 1], 1, 0)) 
 
CLICK APPLY, OK, OK. The BEST RSQUARE variable should now have values of 1, 0. 

The next step would be to select the rows in this data table  where BEST RSQUARE 
has a value of 1. To do this you would go to ROWS MENU, GO TO ROW 
SELECTION, SELECT WHERE, BEST RSQUARE EQUALS 1. This should now highlight the 
rows of the data table where BEST RSQUARE has a value of 1. Now, we can subset 
these rows into a final data table. Go to TABLES MENU, select SUBSET, selected 
rows, OK. The final data table now has the top models for each number of 
parameters.