Model and processing speed
In current desktop and mobile apps, choose the model in Settings → Model. This is one saved account preference. The model label in Chat opens that setting; it is not a second independent preference.
Open it
On desktop and mobile, open Settings → Model. The model label in Chat is also a shortcut to this page.
Choose a model
- Open Settings → Model.
- Choose Auto, or an unlocked named model.
- Use Show all models to see the larger catalog grouped by provider. Use Show fewer to return to the shorter list.
- Wait for Saved. If saving fails, use Retry; the previous saved choice remains in effect.
- Return to Chat and send the next request.
Auto is the recommended starting point. It chooses an appropriate model for the task. Free accounts use Auto; named choices depend on the live plan and catalog. A locked row explains the plan required and opens the upgrade flow.
What the preference covers
Current desktop and mobile Chat use the saved preference when they send a message. Eligible work from email, iMessage/SMS, Telegram, Discord, and scheduled jobs also resolves the account preference. Reopening Model refreshes the saved value, including a change made on another device.
The preference does not promise that every part of a job uses one model. Live audio, image or video generation, specialist tools, and background steps may need their own compatible models. Availability, plan access, and required capabilities can also cause a fallback. A live voice cannot be changed merely by selecting a text model.
If an older app still exposes a separate picker, update it before assuming the controls behave like the current version.
Priority processing
Priority processing appears below the model list. On supported plans it can be switched on for faster eligible replies at a higher cost; other plans show an upgrade option. Read the displayed terms and usage ledger rather than treating it as a guarantee that every tool or network request will finish faster.
Use standard processing when speed is less important. A more expensive model, long context, repeated tools, or media generation can spend credits faster even when a job takes little time.
Fix a missing or unexpected choice
| What you see | What to check |
|---|---|
| Models did not load | Use Retry, check connectivity, and confirm your sign-in session. |
| A model is locked | Read the plan label and current limits. |
| Your change did not stick | Reopen Model and look for the saved selection; retry a failed save. |
| Another device shows an older value | Check that it uses the same account and reopen its Model page. |
| The result used another model | Check availability and the task's required capability; a generation or voice tool may be separate. |
To report a mismatch, include app versions, the selected label, the time and timezone, and the conversation or request reference shown. Do not include credentials.