> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orisu.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

> How AI models work in Orisu: families, variants, and tiers; picking a model on a node; and how settings and cost follow your choice.

Many nodes generate their output with an AI model: image, video, text, and audio nodes all call a model under the hood. You choose which model a node uses, and that choice shapes what the node can do, which settings it exposes, and what it costs to run.

## Families, variants, and tiers

Models are organized in three levels:

* **Family**: a related set of models from one provider that share an identity, like Flux, Veo, or Claude.
* **Variant**: a specific version you can select, like Flux 2 Pro or Veo 3 Fast. This is the thing you actually pick on a node.
* **Tier**: an optional quality or speed level within a variant. When a variant has tiers, picking a faster tier usually trades some quality for a lower cost, and vice versa.

## Picking a model on a node

Select a node on the canvas to open its config panel, then choose the model from the model selector. Only models that can perform that node's job are offered. A generate image node lists image models, a generate video node lists video models, and so on.

When you switch models, two things change:

* **Settings.** Each model exposes its own set of settings. Switching to a different model can add, remove, or replace the fields in the config panel. See [Model settings](/models/settings) for what these controls do.
* **Cost.** The estimated credit cost updates to reflect the selected model, tier, and settings. You see the new estimate before you run the node.

<Note>
  The config panel is the source of truth for which models, settings, and defaults a given node offers. Availability changes as new models ship and older ones are retired.
</Note>

## Provider access and cost

You don't supply provider API keys for these models. Orisu routes each request to the provider and bills the work in credits, so you manage one balance instead of separate accounts and keys per provider. See [Credits and billing](/guides/credits) for how cost is estimated, reserved, and settled.

## Browse models by type

<Columns cols={2}>
  <Card title="Image models" icon="image" href="/models/image">
    Generate, edit, and upscale images.
  </Card>

  <Card title="Video models" icon="film" href="/models/video">
    Generate and transform video, with or without audio.
  </Card>

  <Card title="Text models" icon="text" href="/models/text">
    Generate and reason over text with LLMs.
  </Card>

  <Card title="Audio models" icon="music" href="/models/audio">
    Speech, music, and sound effects.
  </Card>

  <Card title="Model settings" icon="sliders" href="/models/settings">
    Reference for every setting models expose.
  </Card>
</Columns>
