Hi Retell Support Team,
I’m trying to understand the new surcharge behavior around the 3,500 token threshold. At the moment, the conversation history alone can reach ~3,000 tokens before we even include the system prompt or any injected context, which makes it difficult to stay under the limit in production. Could you please explain how the token surcharge calculation works now—specifically how conversation history is handled and counted toward the 3,500 tokens? Any official guidance or documentation you can point me to would be really helpful. Thanks, Thiago
Hello @thiago,
The surcharge (Rule 2) applies when the agent’s prompt exceeds 3,500 LLM tokens. The token count includes:
-
Global prompt + function (tool) descriptions
-
State / node prompt
-
Transcript between agent and user (your conversation history)
-
Tool call history and results
Calculation: Scaling Factor = Prompt Tokens ÷ 3,500; Billed Duration = Original Duration × Scaling Factor (rounded up). Example: 4,200 tokens on a 60s call → 1.2× → billed as 72s.
So yes — conversation history counts toward the 3,500 threshold alongside system/node prompts and tool descriptions.
For Details you can see ( Exceptions to Our Per-Minute Pricing - Retell AI )
Thank You
