We pretrain and fine-tune efficient small language models, making advanced AI more accessible and practical for everyone to run.