An animal’s ability to respond to its environment relies on effectively decoding neural representations of sensory stimuli. While it is known that primary sensory cortices represent sensory stimuli through specific patterns of activity, it is unknown how these representations are preserved over time. Many primary and higher order cortices have been shown to exhibit changes in neural activity representing the same sensory stimuli even when behavior is stable, a process dubbed ‘representational drift’. This poses a critical challenge for the brain – how can animals maintain stable perception and behavior in the face of dynamic representations? One solution is that specific neuronal subpopulations stably retain their representations over time, while other neurons remain unstable. This would allow the brain to respond dynamically to novel stimuli while maintaining stable representations of familiar stimuli. Here, we test the hypothesis that certain subpopulations exhibit elevated stability using the primary vibrissal somatosensory cortex (vS1) as a model by tracking the responses of individual cortical neurons to the same sensory stimuli over time. Using two-photon imaging, we recorded from cortical neurons while behaviorally stable mice with two whiskers performed an active object detection task. We used an encoding model to classify neurons into subpopulations based on their touch receptive fields. Neurons tuned to whisker touch had either narrow or broad receptive fields, with broader tuned neurons responding to multiple touch types. We find that a sparse subpopulation of neurons with broader tuning show increased stability. Specifically, they were more likely to remain touch responsive and exhibited more consistent responses over time. Overall, our results suggest that a small subpopulation of broadly tuned touch neurons may comprise a particularly important population for stably representing whisker touch, while narrowly tuned neurons may impart the system with flexibility to respond to novel stimuli.