Auto model selection
Let Babbily choose a model and reasoning level for each message.
Model capabilities
Choose a model for speed, reasoning, files, tools, media, context, and API usage budget.
Canvas and artifacts
Create, edit, save, export, and send generated work from a dedicated Canvas.
Voice dictation
Prefer speaking to typing? Use voice dictation in chat or Translate when your device supports it.
Choose a model
Select the model pill in the composer to open the model picker. You have two kinds of choices:- Auto sits at the top of the picker. It chooses a model and reasoning level for each message, so quick questions and demanding work each get a suitable model. See Auto model selection.
- Named models are listed below Auto. Choose one when you want the same model for every reply in a chat, or when you want to set Fast mode or reasoning effort yourself. The first few models appear right away; choose More models or Search models to see the rest.
- Speed.
- Reasoning depth.
- Writing style.
- Context window size.
- File support.
- Tool and media support.
- Fast mode means the selected model can use a faster response path for supported work.
- Sunsetting means the model is still available, but Babbily recommends a newer option for new chats. No models are currently marked Sunsetting.
Adjust reasoning
Choose Manual tool mode and turn on Reasoning. Models with adjustable effort show a Reasoning effort selector beside the microphone. Other reasoning-capable models offer an on/off toggle without a level selector. See reasoning controls by model. Reasoning effort and Fast mode are separate controls. When reasoning is off, Babbily asks the model for its lightest setting; some models, such as Claude Opus 5.5, still reason briefly at their lowest level.Reasoning effort, the Reasoning toggle, and Fast mode are hidden while Auto is selected in the model picker. Auto sets the reasoning level for each reply. Choose a named model to set these controls yourself.
Use Fast mode
Fast mode can be useful for drafting, brainstorming, coding, and iterative work where speed matters.1
Select a supported model
Open the model picker and choose a model with the Fast mode lightning icon. Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna, and GLM 5.3 Flash support Fast mode.
2
Turn on Fast mode
Choose Fast mode using the lightning control near the composer.
3
Send your prompt
Babbily uses the faster response path while keeping your selected model visible in the chat.
The Fast mode control appears only for supported models. “Flash” in a model name does not mean it supports this toggle. Check Fast mode availability if the control is missing.
/create-image or Manual > Deep research, while using Auto, Claude Opus 5.5, or Claude Sonnet 5.5, Babbily asks you to choose another model for that request. Tools that Babbily chooses in Auto tool mode remain available with all three.
Write better prompts
Good prompts usually include:- The outcome you want.
- The audience.
- Any constraints or examples.
- The format you want back.
- Whether Babbily should use search, files, memory, or connected apps.
Continue a thread
Follow-up messages can refer to earlier replies, files, and context in the same thread. If a conversation gets long, Babbily tracks the context window and may compact older context so the thread can continue.Leave a reply running
Babbily generates replies on its own servers rather than in your browser tab. This applies to every chat, so you do not have to stay on the page while a long reply finishes.- Closing the tab or losing your connection does not cancel the reply. Reopen the chat to find it still in progress or already finished.
- A dropped connection reconnects to the same reply. Babbily returns to the reply already in progress instead of starting a second one, so you should not see a duplicate answer.
- Reopening a chat does not redo the work. Babbily restores the saved reply rather than repeating the searches, tool calls, or media generation from that turn.
Temporary Chat replies are generated the same way, but the conversation is not saved, so you cannot return to one after you close the tab.
Stop a reply
While Babbily is writing, the send button in the composer becomes a stop button. Select it (its label is Stop generating) or press Esc to end the reply. The button shows Stopping… until the reply ends. Whatever Babbily already wrote stays in the thread, so you can read a partial answer instead of losing it. Stop also works in the moment right after you send, before any text appears. Stopping does not undo work that already finished. Completed searches, tool calls, and generated files remain part of the turn, and any API usage budget already spent is not returned. To take the thread in a different direction after stopping, send a new message or edit your earlier message.Edit an earlier message
Use Edit message on one of your own messages, revise the text, and choose Send. Babbily generates a new reply using the selected model and current generation controls. With Auto, the edited message is matched to a model again, so the new reply can come from a different model. Wait for the thread’s current action to finish before editing. Original attachments must still be available and supported by the selected model. If an attachment is missing, send a new message with the file attached. If Babbily says the edit may have been saved, refresh the thread before trying again.Refer to an earlier chat
In Auto tool mode, Babbily can look up eligible earlier chats when you ask about a previous discussion. This requires memory to be enabled and a model that supports it. It is a limited search, so give a topic or other identifying detail and review any linked conversation. See past-chat lookup and its boundaries.Get notified when a run finishes
For longer chats or research-heavy prompts, Babbily can notify you when a background run finishes. After your first prompt, a Get notified when Babbily is done card can appear in the bottom-left corner, or above the composer on a phone. Choose Turn on notifications, then allow notifications when your browser or device asks. To skip it, close the card. You can also turn Chat completion notifications on or off anytime in Settings > Account.Completion notifications are optional and depend on browser or device notification support.
Work in Canvas
When a reply becomes a document, table, code snippet, image brief, or structured output, Babbily may open it as a Canvas artifact. Canvas gives you a side workspace where you can edit, save versions, export files, and continue prompting with the latest artifact state included as context. Use Canvas when you want to refine generated work instead of copying it into another app first.Use chat history
Saved chats appear under Recents in the sidebar and in Library. To rename or delete a chat, point to it in the sidebar and select … (More actions), or right-click it. On a touch screen, press and hold it. You can also select the chat’s title at the top of an open thread. To share a chat, select Share at the top right of the thread. You can also add or move your own eligible threads into a project from the same menu.When to start a new thread
Start a new thread when:- You switch topics.
- Earlier context is no longer useful.
- The context indicator is near the limit.
- You want a cleaner answer without old assumptions.
Prompting patterns
Write clearer prompts with goals, context, audience, constraints, and output format.
Known limits
Understand what can vary by model, tool, file, plugin, plan, usage, and context.