> For the complete documentation index, see [llms.txt](https://docs.ideogram.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ideogram.ai/create/models.md).

# Models

Compare Auto, Ideogram 4.5, 4.0, 3.0, the Transparent models and partner models in the Image app.

Pick a model from the **Model** picker in the Image app. Each model has its own settings, and the panel changes when you switch. The list comes from Ideogram's servers, so what you see can change over time.

<figure><img src="https://1799634369-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FzjhNby3LLsIikYuvxAJP%2Fuploads%2Fgit-blob-910d410ffdf8f1fd4b016b7f866c135307253403%2Fmodels-1.png?alt=media" alt="Model picker open with the filter chips My models, Team models, Ideogram, OpenAI, Google and Tongyi highlighted above the All models list"><figcaption><p>Use the chips to filter the list by owner or provider.</p></figcaption></figure>

## Auto

**Auto** sits at the top of the picker, and the Image app opens on it. You write the prompt; Auto picks a model for you when the run starts. If you attach reference images, it picks a model that treats your prompt as an edit instruction.

Auto keeps the panel short:

* **Reference images**, up to 10
* **Prompt**
* **Aspect ratio**: Auto, 1:1, 4:3, 3:4, 3:2, 2:3, 16:9 or 9:16
* **Number of images**: Auto gives 4, or 2 when you attach references
* **Seed**, under **Advanced settings**
* **Add to Collection**

Auto has no Magic Prompt or Quality setting. Because the model isn't known until the run starts, the **Generate** button shows an estimated cost with a `~` in front. You pay what the chosen model would charge.

## Ideogram models

Every plan, free included, can use Auto and all Ideogram models.

* **Ideogram 4.5** is the newest model. It's the only Ideogram model that takes reference images, so use it to edit or combine images with a prompt.
* **Ideogram 4.0** offers very wide aspect ratios such as 23:9 and 3:1, at 1K or 2K.
* **Ideogram 3.0** is the only model with style presets, style reference images, negative prompts and color palettes.

|                   | Ideogram 4.5                                                                           | Ideogram 4.0                                  | Ideogram 3.0                                                                 |
| ----------------- | -------------------------------------------------------------------------------------- | --------------------------------------------- | ---------------------------------------------------------------------------- |
| Prompt            | Text or [JSON](/prompting/json-prompting.md), up to 10,000 characters                  | Text or [JSON](/prompting/json-prompting.md)  | Text                                                                         |
| Reference images  | Up to 5                                                                                | None                                          | None (style reference images instead)                                        |
| Aspect ratio      | Auto or a ratio, then **Resolution** 1K or 2K. With references, **Size** replaces both | Auto or a ratio, then **Resolution** 1K or 2K | 15 ratios from 1:3 to 3:1, default 1:1. No resolution choice                 |
| Quality options   | Very low (with references only), Low, Medium, High                                     | Turbo, Default, Quality                       | Turbo, Default, Quality                                                      |
| Magic Prompt      | On/off toggle. Hidden when you attach references                                       | On/off toggle                                 | On/off toggle                                                                |
| 3.0-only settings | No                                                                                     | No                                            | Style, style presets, style reference images, negative prompt, color palette |

[Image Settings](/create/image-settings.md) explains each setting.

### Transparent models

**Ideogram 4.0 Transparent** and **Ideogram 3.0 Transparent** make images with a clear background. Neither takes reference images. See [Transparent Backgrounds](/create/transparent-backgrounds.md) for other ways to get one.

|                 | Ideogram 4.0 Transparent                                                                       | Ideogram 3.0 Transparent        |
| --------------- | ---------------------------------------------------------------------------------------------- | ------------------------------- |
| Prompt help     | **Prompt upsampling**: Auto, On or Off                                                         | **Magic Prompt** toggle         |
| Size            | **Output resolution** 1K, 2K, 4K or 8K. Ideogram upscales anything above 2K after it generates | **Upscale factor** 1x, 2x or 4x |
| Aspect ratio    | Auto, or 1:1 up to 4:1 and the portrait mirrors                                                | Same list as Ideogram 3.0       |
| Quality options | Turbo, Default, Quality                                                                        | Turbo, Default, Quality         |

Both also have **Seed** and **Add to Collection**.

## Partner models

The Image app also runs models from other providers. Premium models need a paid plan or bought credits; on a free plan, picking one opens an upgrade window.

* **GPT Image 2.0**: up to 16 reference images; Resolution 1K, 2K or 4K. Premium.
* **GPT Image 2.5 Flare** and **GPT Image 2.5 Sunburst**: up to 16 reference images, and you can paint a mask on a reference to limit the change to that area. Adds **Quality** (Auto, Low, Medium, High) and, under **Advanced settings**, **Background** (Auto, Transparent, Opaque). Premium.
* **Nano Banana 2** and **Nano Banana Pro**: up to 10 reference images; Resolution 1K, 2K or 4K. Premium.
* **Z-Image**: no reference images; **Inference steps** 20, 30, 40 or 50 ("More steps trade speed for detail."). Free.
* **P-Image-Ideogram**: no reference images; **Prompt upsampling** on or off, Resolution 1K or 2K or custom dimensions, and **Quality** from Very Low to High. Free.

Partner models have no Magic Prompt toggle. All of them have **Aspect ratio**, **Seed** and **Number of images**.

## Custom and community models

A custom model learns a look from images you upload. Yours appear under **My models** in the picker, and your team's under **Team models**. Select **Browse community models** at the bottom of the picker to use models other people have published.

When you pick a custom model, the panel shows the Ideogram 3.0 settings, since custom models are built on Ideogram 3.0. See [Custom Models](/create/custom-models.md) to train one.

## Older Ideogram models

Ideogram 2.0 and 2a may still appear in the picker. For new work, use Ideogram 4.5 or 4.0.
