In this paper, a novel hybrid‐scaling rule for GaN high‐electron‐mobility transistors (HEMTs) scalable large‐signal model with self‐heating effect is proposed. Due to the complex parasitic characteristics of large gate‐width GaN HEMT devices, nonlinear scaling rules are used to consider the dependence of the extrinsic inductances, extrinsic resistances, and thermal resistance on the gate width. In addition, a drain‐source current model and gate capacitance models are proposed to improve the accuracy of traditional Angelov model. A one stage resistor‐capacitor network is attached in the current model to characterize self‐heating effect. Therefore, a scalable large‐signal model based on a hybrid‐scaling rule is presented for the first time. Compared with traditional linear scalable model, the proposed model has a good prediction for multi‐bias S‐parameters and direct current (DC) I‐V curves of different size devices, and can also significantly improve the fitting degree of S22 parameter at high frequency. Moreover, the proposed scalable model shows a good prediction of large‐signal characteristics for large size GaN HEMT devices.