Choose a model

Select the model name at the top of the thread to open the model picker, choose a group, then choose a model. The next message you send uses it.

The button shows the model you are on now. Before a model resolves it reads Select model.

Pick a model

1. Select the model name at the top of the thread.
2. The first pane is headed Models. It starts with Recommended, then one row per provider. Self-hosted models are grouped under Self-hosted.
3. Select a row to open its models. Recommended splits into Smart and Fast. Use Back or All models to return to the first pane.
4. Select a model. The picker closes and the button shows the new name.

The picker has no search box. Which models appear depends on what your Account allows, so two people can see different groups.

Choosing a model from a provider group makes it your default: new threads open on it until you choose another. Choices made inside Recommended are not kept as the default.

Set how hard the model thinks

Some models expose one setting, Reasoning. A higher level makes the model spend longer before answering and costs more.

1. Open the model picker and select the settings control on the model's row.
2. Choose a level under Select reasoning.
3. Use Reset to default to go back to the level the model ships with.

A model with no settings shows This model doesn't support any customizations.

Nothing else about the model is exposed. There is no control for temperature, response length, or web access; a model that can read the web does so on its own when the question needs it.

If it did not work

• The button reads Select model and no models are listed: see No models are available for your account.
• The model you want is not in any group: your Account does not allow it. Ask your Account administrator.
• A file was refused after you switched model: models accept different file types. See A file was not uploaded.
• Replies are slower than you expect: lower the Reasoning level.