The biggest failure mode here is mixing up a higher model count with actual progress or utility; numbers alone don't tell you much.
Tracking 668 models on OpenRouter by 2025 doesn't mean they're all useful, or even that they work well for anything beyond basic tasks.
It's like having a shed full of cheap wrenches – the count is up, but if they snap when you put any real torque on them, you're not actually better off.
The weakest link is assuming quantity translates to meaningful expansion without testing the real-world performance of these new options.