A control portion of an inkjet recording apparatus is capable of executing an immersion cleaning mode of performing cleaning of an ink ejection surface by immersing the ink ejection surface in a cleaning liquid stored in a closed space between the ink ejection surface and a cap and heated. In the immersion cleaning mode, based on a lapse of time from an end of ink ejection and an ambient temperature, the control portion predicts a solidified state of residual ink remaining on the ink ejection surface and, based on the predicted solidified state, sets a cleaning condition.