The primary goal of local AI inference is fitting a model onto hardware while preserving its quality.