Hey, thanks for the report. This behavior is expected, but yeah, the number is still unpleasantly surprising.
On the legacy request-based plan, new frontier models like Grok 4.6 aren’t billed as one message equals one request. They’re billed by actual token usage (token-based / Max Mode), which then gets converted into requests at a rate of $0.04 per request. More details for legacy plans are in this thread: Legacy Individual Plan - Max Mode Required
Why one session turned into hundreds of requests:
- Grok 4.6 has the Fast toggle on by default, and it’s billed at about 2x the normal mode.
- Long agent sessions reread the full, growing chat context on every tool call, so even small steps near the end of a long session can cost a lot.
How to reduce usage:
- Turn off Fast in the model picker.
- Set a lower reasoning effort.
- Start a new chat when you switch tasks so the context doesn’t keep growing.
- Composer is still the most request-efficient option on your plan.
Let me know if any of this doesn’t match what you’re seeing.