It's all well and good that OpenRouter tracked a rapid increase in models, from 63 to 668, but this glossy talk about "informing and populating" research datasets sounds like pure hype.
Just counting models doesn't mean they're magically ready for research; you still need to process, validate, and integrate that data, which is where the real work happens.
For example, I can count how many new educational apps pop up each year, but that doesn't mean they're all useful for my class without me spending hours testing each one to see what actually changes for my students.