I received this question by e-mail:

I have noticed one thing when running SLAC and FEL using the datamonkey cluster, that is , sometimes the model specified (found based on the model selection tool) does not seem to match the model that SLAC and FEL uses??? For one data set the model selection tool suggested that 001121 was the best model. I specified this model in the substitution matrix. But when SLAC and FEL analyses ran it showed that I was using MG94 X (112242)- this was also shown on the result page? Shouldn't SLAC and FEL show MG94 X (001121) as the model used?? However, when I used the 010010 (HKY) model for the same data set but different tree (model found based on the model selection tool), the SLAC and FEL runs indicated that I was using MG94 X (010010)which I also specified in the substitution matrix.
Am I misintepreting something here ??

001121 is the same as 112242.

The encoding for the model is as follows
Position 1: AC rate parameter
Position 2: AG
Position 3: AT
Position 4: CG
Position 5: CT
Position 6: GT

If you look at the model string and see two position match, then the corresponding two rates match too.
What character is in the string is not important. Both models encode a 3 rate model.

AC=AG (=1 for identifiability; AG is always the '1' rate)
AT=CG=GT
CG

HTH,
Sergei