ProFam-1 is a 251M-parameter autoregressive protein family language model (pfLM), trained with next-token prediction on concatenated, unaligned protein sequences drawn from the same family.