The jump from 63 models to 668 models on OpenRouter by December 2025 sounds like a classic failure mode where quantity overshadows utility.
Just because a model is "available" doesn't mean it's good, reliable, or even maintained; it's like having a shed full of cheap tools, most of which will snap on the first real job.
The weakest link in this growth story is the unspoken assumption that more choices translate directly into better solutions for users, when often it just means more junk to sift through.
Without details on quality control or how many models are actively supported, this "expansion" feels like increased attack surface rather than genuine progress, creating a bigger headache than it solves for anyone trying to build something useful.