Aside from the technical advantages of CLAP by itself, it also can serve as the primary development target, and then be wrapped as a VST2/VST3/AU. At present, most development (which is not using something like JUCE, which, really, is it's own custom format that itself gets wrapped) targets VST2 or AU, and is then wrapped to support the other formats. Most developers target VST2, and then wrap to AU and VST3. CLAP is way more capable of this, because of it's simple function-based API, and its extra technical capabilities.
This is especially important now that new developers are no longer allowed to license VST2, and the handwriting on the wall seems to be saying that in the next 24 months, Steinberg will disallow updates to existing plugins or the development of new plugins on VST2, even for those who already have a VST2 license to do so. This means that every developer who uses VST2 as their primary target will be SOL.
VST3 is a non-starter for primary development because of its limitations and fragmented capabilities (for example, it doesn't even support direct MIDI ins/outs). Steinberg released a half-baked format to the world, and immediately started shutting down VST2. Over time, they have added features to it here and there (like MIDI-learn) but that's not very useful if the DAWs have not yet been updated to support them. So all in all, it's a terrible base format to target.