Response quota in chat, and a remembered model
- New
- Improved
Two chat changes that users notice right away. The composer shows AI responses used against the monthly limit, and hitting the limit shows a note with a billing link instead of a raw error. The model picker now remembers the last model you chose across refreshes and new chats. Every turn also records token counts, and every model has an output cap.
- Responses used and remaining shown in the chat composer
- A clear quota note with a billing link instead of a raw error
- Plans with metered overage are never shown as used up
- The picker remembers your model across refreshes and new chats
- Token counts stored per turn, and every model has an output cap






