# Models

> Every model in the catalog with context window and per-component pricing; each has its own detail page.

One key calls every model below — model ids use vendor names and go straight into the `model` field.

<ModelList lang="en" />

## Choosing a model

| Situation | Suggestion |
| --- | --- |
| Complex planning, long-context reasoning | A flagship model with a large context window |
| Bulk execution, reformatting, extraction | A fast model — often an order of magnitude cheaper |
| Code generation and refactoring | A mid-tier model that is stable in that language; compare on a small sample first |
| Image understanding | Must support vision, see [Vision](/en/guides/vision/) |
| Embedding / reranking | Embedding or rerank models; no output tokens |

## Pricing notes

- Prices are **USD per 1M tokens**, with input, output and cache-read priced separately
- Prices are regenerated from the pricing catalog at every build, from the same numbers pushed to billing
- Discounts, top-up options and the full price list live on
  [www.tokrouter.dev/pricing](https://www.tokrouter.dev/pricing/)

<Callout type="info" title="Models change">
Vendors add, rename and retire models. For a "definitely available right now" list, call
[`/v1/models`](/en/api-reference/models/) instead of treating this page as availability truth.
</Callout>