# Choosing a model

> The two Auto modes, every model family, what the quota multiplier means, routes and fallback, and how to pick the right model for a job.

RoGenie has 107 built-in models from 11 families: Claude, DeepSeek, MiniMax, GPT, Gemini, Kimi, Qwen, Grok, GLM, Mistral, Composer. You do not need any API keys to use them. Pick a model with the **Model** button beside the message box. Each chat keeps the model you last picked in it.

## The two Auto modes

| Mode | What it does | Multiplier |
|---|---|---|
| **Auto (intelligent & cheap)** | The default. Runs a capable, low-cost model at a fraction of the quota, with automatic failover between routes. Handles most wishes well. | 0.058× |
| **Auto (Smartest Available)** | Sends every turn to the smartest model available. The most capable choice, and the priciest of the two. | 0.4× |

If a model is unavailable, Auto routes around it automatically.

## The quota multiplier

The number beside each model is its **quota multiplier**: how fast it uses your quota compared with Claude Opus 5, which is 1× (priced at $5 per million input tokens and $25 per million output tokens). A 0.5× model uses your quota half as fast as Opus 5; a 2× model uses it twice as fast. Auto (intelligent & cheap) runs at 0.058×, and the cheapest model in the catalog right now is GPT 6 Luna (0.02×).

- The label is an estimate for a typical RoGenie run. Under the hood, input and output tokens each have their own multiplier, and you are billed by those exactly. Hover a model's multiplier to see both.
- A model with no published price shows an estimate with `~` and is billed at the default 1×.
- Models that run on [your own API key](https://rogenie.ai/docs/own-api-keys) show **Your API key · no RoGenie quota** and do not use your quota at all.

See [How usage is counted](https://rogenie.ai/docs/usage-and-quota) for the full picture.

## Model families

As of 2026-11-09. The live catalog in the app changes often, so new models may already be there.

| Family | Models | Lowest multiplier | Highest multiplier |
|---|---|---|---|
| Claude | 19 | 0.02× | 3× |
| DeepSeek | 8 | 0.051× | 0.25× |
| MiniMax | 10 | 0.029× | 0.12× |
| GPT | 23 | 0.02× | 2× |
| Gemini | 8 | 0.15× | 0.41× |
| Kimi | 13 | 0.12× | 0.8× |
| Qwen | 11 | 0.023× | 0.47× |
| Grok | 4 | 0.38× | 0.76× |
| GLM | 9 | 0.029× | 0.27× |
| Mistral | 1 | 0.26× | 0.26× |
| Composer | 1 | 0.1× | 0.1× |

Models marked **Recommended** in the picker are good defaults within their family.

## How to choose

| You want | Try |
|---|---|
| Everyday building, small fixes, assets, UI | **Auto (intelligent & cheap)** |
| Hard logic, tricky bugs, large refactors | A stronger model, or **Auto (Smartest Available)**, and a higher [effort](https://rogenie.ai/docs/effort-and-fast-mode) |
| To stretch your quota | A model with a low multiplier. When you reach 75% of a quota window, RoGenie offers to switch you to Auto (intelligent & cheap). |
| A model you already pay for elsewhere | [Add your own key](https://rogenie.ai/docs/own-api-keys) for that provider |
| Different models for different jobs in one run | Pick per role in [Agent settings](https://rogenie.ai/docs/agent-settings#sub-agent-models), for example a cheap model for asset searches and a strong one for scripting |

## Routes and fallback

Many models are offered through more than one **route** (a provider that serves the model). The picker shows how many routes a model has and which one is currently best. RoGenie uses the best available route and **falls back automatically** if it fails, so a model keeps working when one provider has a problem. Routes are shown as *Provider 1*, *Provider 2* and so on.

You do not choose or order the routes yourself: RoGenie picks them for you. To run a provider's models on your own account instead, [add your own key](https://rogenie.ai/docs/own-api-keys).

## Using the picker

- Type to search: a model name or a family like "Claude" or "Gemini".
- Filter by family with the tabs.
- Keyboard: `↑` `↓` to move, `Enter` to select, `→` to see a model's routes.
- A model that is not available right now is greyed out with the reason.

If a chat's model is no longer available when you come back to it, RoGenie switches to another one and tells you.

## Images and models

Not every model can read images. If you attach an image and the model cannot see it, RoGenie tells you; switch to a model that supports images.