Expanded Token Windows with Proportional Message Deduction

It would be beneficial to introduce a manual toggle that allows users to access higher token limits for the more expensive models. This feature should be designed in a way that aligns with your existing usage model, similar to Claude.ai’s approach, where increased token usage is offset by chat consumption.

For example, if the default context window for Sonnet is 8k tokens, enabling an "Unlimited Tokens" toggle would allow the model to process more tokens. However, the additional tokens would be deducted from the user's available chats. The deduction could be calculated by dividing the total tokens used by the standard token limit per chat.

For instance, if a user utilizes 23k tokens in a single chat, it would cost them 3 chats (since 8k × 3 = 24k). This approach ensures flexibility while maintaining fairness in resource allocation.