Model routing

Model routing is the practice of sending each request to a model chosen for that request, rather than using one model for everything. The usual aim is to spend on a strong model only where quality matters and use a cheaper one elsewhere.

What it is

Routers decide in different ways:

• By task type: planning and review go to a strong model, routine drafting to a cheap one.
• By difficulty: a classifier or a cheap first attempt decides whether to escalate.
• By availability: fall back to another model when one is down, rate-limited or out of budget.

Routing interacts with prompt caching: a cache belongs to one model, so switching models mid-conversation starts the cache again.

Why it matters

In an agent session, a few calls steer everything after them, and many others do routine work. Paying top-model prices for every call is the most common way an AI coding bill grows.

In AI CodeOps

The platform's models come in three tiers: Economy, Standard and Premium. A session runs on one tier, or on Auto, which picks a tier for each model call:

planning, review, security    Premium, then Standard, then Economy
writing the change            Standard, then Premium, then Economy
other drafting                Economy, then Standard, then Premium

Each call takes the first tier in its order that your plan includes and still has allowance for. Auto is offered once at least two tiers can be chosen.