### Answer Going over 4,000 LLM prompt tokens triggers proportional scaling of billed call duration rather than a fixed per-model rate increase [(1) ]() . The scaling works as follows: - Scaling Factor = Prompt LLM Tokens ÷ 4,000 - Billed Duration = Original Duration × Scaling Factor (rounded up) Example: a 60-second call using 4,800 tokens has a scaling factor of 4,800 ÷ 4,000 = 1.2, resulting in a billed duration of 72 seconds (12 seconds additional charge) . Token calculation includes the global prompt, functions (tool descriptions) , state/node prompt, transcript between agent and user, and tool call history and results . Flex mode is noted as a common trigger for this rule, since it compiles all node prompts, transitions, and tool descriptions into a single LLM context, often pushing token count well above 4,000 . I don't have per-model breakdowns of this scaling rate in the sources — it applies as a general proportional formula regardless of model. If you'd like, I can look into whether a specific model's base per-minute rate is documented separately.

​ To learn more, see the sources I considered: [[1] Exceptions to Our Per-Minute Pricing](https://docs.retellai.com/accounts/billing-exceptions) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.