Chat basics

Every conversation happens in one place: type in the message box at the bottom and press Enter. This page covers the controls around that box. For your history, editing and branching, see Conversations. For attachments, see Files and images.

Choose a model

Click the model name at the top left of the chat to open the picker. There are two models:

In the picker	What it is
GLM-5.3 Flash	The classical model. Fast, strong at tools and code. The default for every new chat.
GLM-5.3 Flash Unfiltered	The same model, uncensored.

The picker and the Unfiltered switch below do the same thing, and each keeps the other up to date. The choice belongs to the conversation, so switching in one chat doesn’t change the others.

The unfiltered model is often slower to start, sometimes taking half a minute before the first words appear. That’s normal, not a hang. Wait for it instead of sending again, because every resend is a new request.

The Unfiltered switch

Unfiltered sits in the row under the message box. Turn it on and your next messages go to the uncensored model. Turn it off to go back.

If a reply from the standard model hedges or refuses, you don’t need to retype anything. Under the latest reply you’ll see “Not the answer you wanted?” followed by Retry unfiltered. Click it and Aderra turns the switch on and writes that reply again on the unfiltered model. The conversation stays on the unfiltered model afterwards.

The switch appears in normal chats. A chat with one of your saved agents uses the agent’s own model instead, so the switch and the reasoning menu below aren’t shown.

How hard it thinks

Next to the switch is the reasoning menu. It starts on Fast. Open it to choose:

Fast: answers straight away. Right for most questions.
Medium: thinks before it answers.
Thinking: thinks longest. Use it for maths, tricky code or careful planning.

The more it thinks, the longer a reply takes and the more tokens it uses, so it costs more credits. The setting applies to the current conversation. On a phone the menu shows only a gauge icon.

When the model thinks, its reasoning appears above the answer in a collapsible Thoughts section. To have that section open by default, turn on Open Thinking Dropdowns by Default in Settings → Chat.

Writing a message

Enter sends. Shift+Enter starts a new line.
To make Enter add a line instead, turn off Press Enter to send messages in Settings → Chat. Then send with Ctrl+Enter (⌘+Enter on a Mac).
Paste a long block of text (about 2,500 characters or more) and Aderra attaches it as a text file called pasted-text.txt instead of filling the box. Click the file to edit it, or choose Move back into message to put it back into the box as text.
To add a document or an image, drag it onto the chat or paste it into the box. See Files and images.
On a new chat you’ll also see three starter buttons under the box: Explore an idea, Write something and Solve a problem.

The row under the box can also show other chips. For example, one of them controls building apps and pages. The header has a menu for choosing the image model.

While Aderra is answering

To stop a reply, click the square stop button that replaces the send button, or press Ctrl+Shift+X (⌘+Shift+X). If you’ve already started typing your next message, the send button takes the stop button’s place: clear the box to get the stop button back, or use the shortcut. Whatever was already written stays in the chat.

To add something mid-reply, type your next message while the reply is still being written. Pressing Enter usually steers. During the first reply of a new chat, or while a tool is waiting for your approval, steering isn’t possible: Steer is greyed out and Enter queues instead. For another option, hover over the send button (or move the keyboard focus to it):

Steer: adds your message to the current reply at its next step. This is what Enter normally does.
Queue: sends your message as a new turn once the reply finishes.
Interrupt & steer: stops the reply now, keeps what it wrote so far, and carries on with your message.
Interrupt & send: stops the reply and sends your message as a new turn.

Long replies. Building something or working through several steps can take a while. After 30 seconds the chat shows “This can take a few minutes. You can switch tabs.” with a Notify me when it’s done link. Click it and allow notifications in your browser. If the tab is in the background when the reply finishes, you’ll get a notification. Click it to go back to the chat. The tab has to stay open for this to work.

“Reached the tool call limit for this turn”. On a long task Aderra can use every step it is allowed in one turn. Everything so far is saved. Click Keep going to let it continue, or Answer now to get an answer from what it already has.

How replies are shown

Formatting. Replies use Markdown: headings, lists, tables, bold and links are shown formatted.
Code. Code appears in blocks labelled with the language. Each block has Copy code and Download code buttons.
Maths. LaTeX formulas are shown as equations.
Diagrams. Mermaid diagrams are drawn as pictures.
Buttons under each reply. You can copy a reply, rate it, write it again and more. See Working with replies.

Language

The app is available in 51 language options, including English, with two versions each of Chinese and Portuguese. Choose yours in Settings → General → Language, or with the globe at the top right of the sign-in page. Auto detect follows your browser.

The language setting only changes the app’s menus and buttons. Write to Aderra in any language. It usually replies in the language you wrote in.

Keyboard shortcuts

Press Ctrl+Shift+/ (⌘+Shift+/ on a Mac) to see every shortcut. You can also open the list from your account menu at the bottom left, under Help → Keyboard Shortcuts. You can change the keys there too. Some useful ones:

Action	Windows / Linux	Mac
New chat	Ctrl+Shift+O	⌘ ⇧ O
Focus the message box	Shift+Esc	⇧ Esc
Open the model picker	Ctrl+Shift+M	⌘ ⇧ M
Stop generating	Ctrl+Shift+X	⌘ ⇧ X
Regenerate the last reply	Ctrl+Shift+E	⌘ ⇧ E
Edit your last message	Ctrl+Shift+I	⌘ ⇧ I
Copy the last reply	Ctrl+Shift+;	⌘ ⇧ ;
Copy the last code block	Ctrl+Shift+K	⌘ ⇧ K
Temporary Chat on or off	Ctrl+Shift+Y	⌘ ⇧ Y
Show or hide the sidebar	Ctrl+Shift+S	⌘ ⇧ S

Temporary chats

Turn on Temporary Chat before your first message. It’s the dashed speech-bubble button at the top right. On a phone, find it in the More options menu. The chat stays out of your history and is deleted automatically: each message is removed 7 days after it was sent. To start every new chat this way, turn on Temporary Chat by default in Settings → Chat. See Temporary chats for the details.

How long a conversation can be

Each plan has a context window: how much of the conversation the model can read at once. Free covers shorter conversations. Paid plans go much further, and the Plan page lists each plan’s window. Attached documents and images count towards it too.

When a chat grows past your window, you’ll see “This conversation is about …k tokens, over the … context window of the … plan.” Start a new chat. Asking this chat for a summary won’t work, because that request carries the same too-long history. To keep context, copy what you need, or use Fork on an earlier message, from before the chat got too long.

On the Free plan

Free includes 10 messages a day, back at 00:00 UTC. Each message you send counts as one. So does each Regenerate, rerun, Retry unfiltered and Keep going. With two answers side by side, a message counts twice. The steps Aderra takes while answering don’t count, and neither do messages that fail. The Plan page shows Free messages left.

When they run out, the chat shows “You’ve used today’s 10 free messages. They come back at 00:00 UTC…” with a See plans button. To keep chatting today, choose a paid plan or buy a credit pack. Buying credits lifts the 10-message limit. See Plans and credits.

Not in the app

The chat app doesn’t offer web search, long-term memory across chats, sharing a conversation by link, searching your chat history, or running code on a server.