Quality predictive models were developed to predict the freshness of bighead carp (Aristichthys nobilis) fillets during storage at different temperatures. Quality indices [sensory score, total volatile base nitrogen (TVB-N), total aerobic counts (TAC) and K value] at −3, 0, 3, 9 and 15 °C were estimated and kinetically modelled by the Arrhenius equation. The activation energy (EA) of sensory score, TVB-N, TAC and K value was 78.17, 75.93, 106.53 and 76.21 kJ mol−1, and the corresponding rate constants (k0) were 1.16 × 1015, 2.60 × 1014, 4.05 × 1019 and 1.36 × 1015. The high regression coefficients (R2 > 0.87) indicated the acceptability of the zero-order reaction for sensory score, TVB-N, TAC and K value. Relative errors between predicted and observed freshness indicators values of sensory score, TVB-N, TAC and K value were all below 10% except the values at 6th day of K value, 3rd day and 9th day of TVB-N. These results indicated that the models based on sensory score, TVB-N, TAC and K value could effectively predict the freshness indicators of bighead carp fillets at the range of −3 to 15 °C.