This article addresses the problem of fast fixed-time tracking control for robotic manipulator systems subject to model uncertainties and disturbances. First, on the basis of a newly constructed fixed-time stable system, a novel faster nonsingular fixed-time sliding mode (FNFTSM) surface is developed to ensure a faster convergence rate, and the settling time of the proposed surface is independent of initial values of system states. Subsequently, an extreme learning machine (ELM) algorithm is utilized to suppress the negative influence of system uncertainties and disturbances. By incorporating fixed-time stable theory and the ELM learning technique, an adaptive fixed-time sliding mode control scheme without knowing any information of system parameters is synthesized, which can circumvent chattering phenomenon and ensure that the tracking errors converge to a small region in fixed time. Finally, the superior of the proposed control strategy is substantiated with comparison simulation results.