Learn AI · Routing
The invisible intelligence behind
every good AI answer.
When you type a prompt, which model answers? The router decides. Here is how routing works, in plain language.
What model routing is
A model router is the decision layer that sends each request to the best available model — based on your plan, the task, the model's cost, and its current availability. Instead of picking one model and living with its weaknesses, a router picks per request.
You never see the machinery. You just get better answers, faster, at a lower price.
The ssv.asia lanes
| Route | Pool it meters | Best for |
|---|---|---|
ssv.asia/free | Free tokens | Everyday questions, drafting, rewriting |
ssv.asia/pro | Pro tokens | Long refactors, deep analysis, priority speed |
ssv.asia/aec | Pro tokens (AEC pass) | Architecture/engineering with cited RAG |
ssv.asia/foundation | Foundation tokens | Heaviest frontier work (+ Opus boost) |
You pick the lane by how you work; the router picks the model inside it. See Frontier models explained for who is behind the lanes.
Why routing beats one model
- Right tool per task — a one-line question and a 2,000-line refactor deserve different models.
- Cost control — cheap models for routine work, premium only when it matters.
- Resilience — one lab's outage does not stop your work.
- No lock-in — new leaders are adopted automatically.
Want to go deeper on any term? The glossary has you covered.