Choose a chat model and thinking level

Purpose

A quick reply to a customer and the month's sales analysis do not need the same effort. Neda picks the model that answers her next message and how long it thinks first, so simple questions stay fast and hard ones get more care.

Before you start

Open a chat or a new chat. The models you see are the ones your plan and workspace allow. Availability can change, so the list shown is always the current one.

Steps

1. Look at the end of the message box. The model button shows the model's name and, for models that can think, the current level, for example "Low thinking". Select it.
2. If the model can think, a small panel opens first. Drag the slider to set the level: "No thinking", "Minimal thinking", "Low thinking", "Medium thinking", "High thinking" or "Extra high thinking" (the model offers only the levels it supports). A line under the level explains it, for example "Balanced speed and depth". Higher levels give deeper answers but use your allowance faster, and a warning marks them.
3. To switch models, select the model name in that panel ("Change model"). Models that cannot think open the list directly.
4. In the list, type in "Search models" to filter. With an empty search the list shows your "Favorites" first, then a short list of recommended models, with "Auto" among them. Select "More models" to see every model, grouped by maker. Select a model to use it.
5. Read the badges on each row. "Vision" means it can read pictures. The brain badge ("Reasoning") means it can think. "Tools" means it can use web search and plugins. "No tools" means chat only: it cannot use tools and is not offered in Work mode. "Auto" shows a short hint about what it does, and the most expensive models carry a warning ("Uses your quota fast"). A shield marks a private model, one that runs only on servers inside Iran. Where Data Privacy Mode applies, the other models are listed greyed out under "Not available in Data Privacy Mode" with "Not hosted in Iran" and cannot be chosen (see Use Data Privacy Mode).
6. Select the star at the end of a row ("Add to favorites") to pin a model, and select it again ("Remove from favorites") to unpin it. Drag a favorite by its grip ("Drag to reorder") to change the order. Your first favorite carries the "Default" tag: new chats start with it.
7. "Auto" lets Agentum pick the right model for each message. Where Data Privacy Mode applies, "Auto" always uses a private model. The starting selection of a new chat follows the order described under "Expected result".
8. Send your message. The choice applies from this message on, and the chat remembers it when you come back.

Expected result

Your next message is answered by the model and thinking level you chose. For a new chat the starting model is, in this order: the one you picked, the first usable favorite, the last model you used in this browser, then your default language model from Settings, "Models" (which is "Auto" unless you changed it). For a custom agent, its configured model is the fallback; for the default agent, your default language model is used. A saved chat starts from the model it last used. Default models for each kind of task (language, images, speech) are set in Settings, "Models". In Work mode only models that can use tools are listed.

Troubleshooting