What's this about?
Cursor, the AI code editor, launched "Router" — a system that automatically picks which AI model should handle each coding task instead of making you choose. Cursor claims it delivers frontier-model quality at 60% lower cost.
What exactly is a "model router"?
Different AI models have different strengths and prices — like specialists in a hospital. A router is the triage nurse: it looks at each incoming task and sends it to the cheapest model that can do the job well, saving the expensive specialist for the hard cases. Yesterday's Fireworks study showed routing between Kimi K3 and Fable 5 beats either model alone; Cursor is now productizing the same idea.
What's the community saying?
7.4k likes on the announcement; no big HN thread yet
- It fits a visible industry turn: Cursor's own SQLite-rebuild experiment a day earlier showed a 15x cost spread depending on model mix — model choice is becoming an optimization problem, not a preference.
- Skeptics of routers (including Ethan Mollick this week) argue users need *more* control over which model does what, not less — routing hides the decision.
- The 60% number is Cursor's own; watch for independent benchmarks.