Running an AI model on standard personal computers or laptops rather than expensive specialized servers or cloud infrastructure.