Frozen core optimization

Dear all,
I would like to optimize large cluster (like Boron -12). If I freeze the core, how much it would effect the final result ? Or should I use lower basis sets and method instead of using frozen core approximation ?

Sometimes gradients aren’t implemented for set freeze_core true, so you may find more method flexibility without fc. Note that you may wish to use core-inclusive basis sets like cc-pCVXZ when enabling core.