Choose a model and reasoning level

Each Glean Tau chat uses one model at a time, along with a reasoning level that controls how much the model thinks before it responds. The model button in the composer shows both, such as the model name followed by High.

Understand reasoning levels

Higher reasoning levels give the model more time to work through a problem before it answers. They usually help with multi-step tasks, such as debugging, planning a change across several files, or analyzing a long document. They also take longer and use more tokens, which increases cost. Lower levels respond faster and cost less, and work well for quick questions and simple edits.

Level	Use it for
Off	Quick answers where speed matters more than depth
Low	Simple edits, lookups, and short summaries
Medium (Med in the picker)	Everyday tasks that need some reasoning
High	Multi-step work. This is the default for new chats.
Extra high (XHigh in the picker)	Difficult problems where accuracy matters more than speed
Max	The hardest problems, on models that support it

Not every model supports every level. You can choose only the levels the selected model supports, and the keyboard shortcut and /reasoning command skip the rest.

To see how reasoning affects cost, check Track model cost and usage.

Change the model or reasoning level for a chat

Changes apply to the current chat only. Glean Tau confirms the change in the chat, such as "Reasoning changed to medium."

Model picker

1

Open the model picker

Select the model button in the composer, or press Cmd+/ (Ctrl+/ on Windows).

2

Pick a model

Search or browse the model list, then select a model.

3

Pick a reasoning level

In the row of reasoning levels below the model list, select a level, such as Med or High. Hover a level to see its full name.

Keyboard shortcut

Press Cmd+Shift+/ (Ctrl+Shift+/ on Windows) to cycle through the reasoning levels the current model supports. To change the shortcut, see Keyboard shortcuts.

Slash commands

• Type /model to open the model picker. Add search text, such as /model sonnet, to filter the list.
• Type /reasoning to open the picker, or add a level, such as /reasoning low, to switch directly.

The picker lists the models your organization makes available, minus any you've hidden. If a model you expect is missing, check Choose which models appear. If it isn't there either, ask your admin. Admins choose models in the Model Hub, as described in Configure models.

If you attach an image, models that don't support images are unavailable in the picker. Remove the image to use one of those models.

Set the defaults for new chats

Go to Settings → Models. Under New chats, set:

Setting	What it controls
Default model	The model and reasoning level used when you start a new chat. Set the reasoning level in the picker's row of reasoning levels. The default level is High.
Start new chats with	Whether new chats use your default model (Default model) or carry over the model you last chose (Last used model)

These settings don't change chats that are already open.

Choose which models appear

Go to Settings → Models, or select Manage models at the bottom of the model picker, to keep the picker short. Under Your model list, turn off the switch next to any model you don't use to hide it from model pickers across the app. Use the search box, or filter by All models, Enabled, or Hidden, to find a model. Each model also shows its capabilities, such as Image input.

You can't hide your default model, and at least one model must stay visible. Hidden models stay available in chats that already use them, and you can show them again at any time.

See also

• Agent approval modes — control what the agent can do without asking