> For the complete documentation index, see [llms.txt](https://anyint.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://anyint.gitbook.io/docs/smart-routing-coming-soon/what-is-anyint-smarting-routing.md).

# What is Anyint Smarting Routing

Smart Routing is an important part of the AnyInt product direction.

It is designed for a simple problem: most users know the task they want to complete, but they do not know which model is the best fit. As a result, many teams default to expensive premium models even when a more affordable model could achieve the same practical outcome.

Smart Routing is not fully available yet, but it represents a core part of how AnyInt will help users get better results with less manual trial and error.

### Why it matters

In real-world usage:

* the same task can often be handled by more than one model
* the difference in output quality is not always large
* the difference in price can be very large
* most users care about the outcome, not the model brand

Smart Routing is intended to reduce that decision burden and help users choose a model that is more appropriate for the task, budget, and performance target.

### What Smart Routing is meant to optimize

* quality when the task is difficult or high-stakes
* cost when a lighter model can achieve the same useful result
* speed when low latency matters more than maximum intelligence
* consistency when teams want better defaults across workloads

### Example situations

* simple extraction or classification tasks that do not need a premium model
* everyday chat or assistant tasks where response speed matters
* code, reasoning, or long-context requests that may justify a stronger model
* mixed workloads where one fixed model is either too expensive or too weak

### Product direction

Over time, Smart Routing is intended to help users:

* avoid overpaying by default
* match requests to more suitable models automatically
* combine routing, fallback, and policy into one decision layer
* move from manual model guessing to more informed model selection