4 ms·

I'm not sure if you're aware of the amount of text used to train these large language models, but in that context, yes it is miniscule.