Introduction: The growing need to explore a broader range of cardiac electrophysiological conditions has led to increased simulation complexity and, consequently, higher computational costs. To address this, we present Cocoro, a fully GPU-resident solver powered by the cutting-edge WebGPU standard for efficient computation and rendering. Methods: We thoroughly describe our implementation and benchmark its accuracy, performance, and clinical potential. Our results were compared with those of the gold standard simulator openCARP (OC) in terms of activation times (ATs) and action potential duration at 90% repolarization (APD90) when a porcine high resolution biventricular mesh (BiVh) was used and both anisotropic (A) and isotropic (I) setups were considered. To assess efficiency, the BiVh and its low-resolution version (BiVl) were used for measuring Cocoro's execution time (ET) for 5-second simulations on two GPUs with and without extra graphical information (EGI), including visualizations of pseudo-electrocardiograms and cellular data. In addition, we compared the simulated and experimental QRS complexes to showcase clinical applicability. Results: Cocoro simulations were closely aligned with those of OC, with the 90th percentile (P90) of node-wise AT differences being 11 ms (A) and 4 ms (I). For the APD90, P90 remained below 6 ms for both A and I. On an RTX GPU, a 5-s simulation took 2.9 min for the BiVl mesh and 9.5 min for the BiVh mesh. On a Titan GPU, ETs were 6.3 and 22.1 min for the BiVl and BiVh meshes, respectively. Enabling EGI had a minimal impact on ETs. Moreover, the simulated QRS complexes reproduced the experimental QRS morphologies and durations. Conclusion: Thus, Cocoro allows fast, portable, and accurate fully GPU-resident cardiac electrophysiological simulations.