Affirm (AFRM) said Thursday it has launched a transformer-based machine-learning model for credit underwriting in the US.

The model analyzes and learns from the order and timing of events in a consumer's credit history, the company said.

Its initial deployment, the model approved applications that would otherwise have been declined, which resulted in a 3.4% increase completed purchases and loans that performed better than a comparable expansion under a previous machine learning model, Affirm said.

The new model is live at checkouts in the US, the company said.

Affirm shares rose 5.6% in premarket trading.