<p>The development of a highly accurate predictive model for ultra-fine austenite grain and the delineation of a heat treatment process window have emerged as the primary focus of interest for steel companies. In this paper, a new type of austenite grain growth model with excellent prediction accuracy for ultra-fine austenitic grains was proposed by using a new twice error fitting solution method. And the formation mechanism of ultra-fine austenite grains was analyzed by austenite nucleation and second-phase pinning action. The results indicated that flattening of prior austenite grain could promote the nucleation of austenite grains during the reverse transformation process. Combination with the pinning effect of nano-sized V(C, N) particles resulted in the formation of extremely fine austenite grains. In the reheated temperature window of 820–880&#xa0;°C, ultra-fine austenite grains of less than 5&#xa0;μm could be obtained within 10&#xa0;min, exhibiting a strong resistance to coarsening when insulated. Subsequently, a new type Sellar austenite grain growth model with two stages was established based on the growth characteristics of austenite grains and the pinning force of V(C, N) particles. After error accounting, the application of new twice fitting method had resulted in a significant reduction in error and an obvious improvement in accuracy. The average absolute relative error (<i>E</i>_AAR) and root mean square error (<i>E</i>_RMS) of the new type Sellar model significantly decreased from 15.7% and 9.83&#xa0;μm to 2.51% 0.03&#xa0;μm.</p> Graphical abstract <p></p>