▶ Stepthrough Courses All tutorials Blog Glossary Prompts Videos Visual guides Cheat sheets Comparisons Start Learning Free

What is a model router?

Quick answer

A model router is software that decides which AI model should handle each request. Easy questions go to a fast, cheap model; hard ones go to a stronger or reasoning model. Some chat apps route for you automatically, and automation builders can set up their own routing rules.

Last updated

Updated · By Robert Breen

Why it matters for a small business

Routing is why you often don't pick a model at all. ChatGPT, for example, offers modes such as Instant and Thinking and can choose how long to reason, so a quick rewrite comes back fast while a tricky spreadsheet question gets more thinking time. Perplexity and Microsoft Copilot also choose models behind the scenes on many plans. The upside is speed and lower cost; the downside is that the same question can be answered by a different model on different days.

In your own automations, a simple router saves real money. A first step uses a small model to sort each incoming email, and only the messages that need a careful reply go to a larger model. Because token pricing between tiers can differ by 50 times or more, routing the easy majority to a small model can cut the bill sharply. Keep the rules simple and log which model answered, so you can check quality later.

In a real lesson: Gmail AI Triage: Auto-Label Every Email and Draft Replies

The Gmail AI Triage lesson builds the pattern a router uses, even though it calls one model. The workflow sorts every incoming email into a label first, then drafts replies only for the messages that need one.

A router adds one more decision at that fork: the sorting step can run on a small, cheap model, and the drafting step can use a stronger one. In n8n that means a different model in each step, chosen in each node's Model dropdown.

The n8n inbox triage workflow: Gmail Trigger, a Text Classifier with Lead, Client and Invoice outputs wired to three label steps, then a Basic LLM Chain and Create a draft on the Client branch
The n8n inbox triage workflow: Gmail Trigger, a Text Classifier with Lead, Client and Invoice outputs wired to three label steps, then a Basic LLM Chain and Create a draft on the Client branch

Try this lesson free or read the step-by-step guide.

Common confusions

Model router vs AI agent

An AI agent decides which tools to use to finish a task. A router only decides which model answers.

Model router vs model family

A model family is the set of models to choose from. The router is what does the choosing.

Tips

  • Route by simple rules first, such as message length or label, before anything clever.
  • Record which model answered each request so you can audit quality.

More AI basics terms

Where you use it: free lessons

Frequently asked questions

Does ChatGPT use a model router?
It offers modes such as Instant and Thinking and can adjust how much it reasons for each request, as of October 2026. You can still choose a mode yourself on paid plans when you want a specific behavior.
Is a model router worth it for a small business?
In chat apps it is built in, so you get it free. In your own automations, it is worth adding once a workflow runs often enough that the API bill matters.

All AI glossary terms, A to Z · Free prompt templates