Model Predictive Torque Control (MPTC) has been widely employed in induction motor drives owing to its fast dynamic response and straightforward implementation. However, conventional MPTC suffers from two major limitations: the need for weighting factors and its dependence on an accurate machine model. Sequential Model Predictive Control (SMPC) has emerged as one of the simplest and most effective strategies by replacing the conventional multi-objective cost function with a sequence of single-objective cost functions. Nevertheless, conventional SMPC still relies on an accurate motor model and employs a fixed priority order for the cost functions, which may lead to suboptimal voltage vector selection under different operating conditions. To address these limitations, this paper proposes an adaptive model-free sequential predictive torque control strategy for induction motors. An ultra-local model combined with an Extended State Observer (ESO) is employed to estimate the system dynamics online, eliminating the dependence on machine parameters. Furthermore, an adaptive cost-function prioritization strategy is developed, in which the normalized torque and flux tracking errors are evaluated online to determine the order of sequential optimization. Simulation results demonstrate that the proposed method controls torque and flux effectively and improves robustness against parameter mismatches.