### Answer To reduce token usage in Retell AI, focus on keeping your prompt length under 3,500 tokens. Here's why this matters: Understanding Token-Based Billing Retell AI scales your LLM pricing when prompts exceed 3,500 tokens. [(1) ]() What counts toward your token calculation: - Global prompt - Functions (tool descriptions) - State/node prompt - Transcript between agent and user - Tool call history and results Price Scaling Formula - Scaling Factor = Prompt LLM Tokens ÷ 3,500 - Billed Duration = Original Duration × Scaling Factor (rounded up) For example, a 60-second call using 4,200 tokens would be billed as 72 seconds (4,200 ÷ 3,500 = 1.2 scaling factor) . Practical Tips to Reduce Tokens 1. Keep prompts concise – Trim unnecessary instructions from your global and node prompts 2. Minimize tool descriptions – Use brief, clear descriptions for functions 3. Limit tool call history – Reduce the number of tools or consolidate where possible Knowledge Base Usage Based on community discussion, knowledge base queries only incur token usage when the knowledge base is actually called, so structuring your agent to call it selectively can help reduce average token use per call. [(6) ]()

​ 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.