The answer to this question depends on what process one is interested in simulating well. As you know, short/mid-range weather prediction models and climate models have very different applications and goals.
Because the short range weather prediction model is typically of much higher resolution than climate models (~1-10 km versus ~50-200 km), it is almost always more skilled at simulating clouds forming at a specific location and time. Cloud and rain prediction skill tends to be greater in synoptic scale fronts and near topographic features, and smaller for sporadic, small-scale, tropical thunderstorms. In very high-resolution NWP models (2-3 km or smaller), convection and clouds may be simulated reasonably well without using any cloud parameterization scheme. Nevertheless, clouds and rainfall are still hard to simulate or forecast very accurately.
On the other hand, due to very low grid resolution, climate models can really aim only for correctly simulating the occurrence frequency and the amount of clouds and rain in a larger geographical region over a longer period of time. In climate, clouds play a significant role in radiative feedbacks. Climate models still rely on cloud parameterization schemes.