This is unfortunately a known error with the new rstan 2.26, where if the parser encounters an error, then all future models (even if correct) will also error. The only solution at this point is to restart the R session.
@hsbadr did we have a workaround for this yet?