Call duration and telephony charges
The agent’s call and the phone connection can end at different times. For Retell-managed numbers, telephony charges can continue after the agent leaves. With custom telephony, your provider bills the phone connection; Retell does not add a telephony fee.
For calls that end without a transfer or takeover, duration-based charges use the recorded call duration. Normal hangup and error cleanup do not add a separate billable interval. The billing adjustments below can still change the billed usage.
For transfers that stay connected through Retell-managed telephony,
call_ended and call_analyzed can arrive before the final telephony charge. Use the updated cost in transfer_ended, or retrieve the call after the transfer ends. See transfer billing webhooks.
Calculate the cost
Telephony cost = connected seconds ÷ 60 × the applicable per-minute telephony rate. For transfers and fallback calls, the rate depends on the provider and destination. Check the pricing page for country rates and your workspace’s pricing for any contracted discounts. The examples below use $0.015/min for telephony, the published US example rate, and $0.11/min for agent usage before telephony (as of November 2026). The agent rate combines $0.055/min voice infrastructure, $0.04/min GPT 5.1 on the standard tier, and $0.015/min Retell platform voices. See the published cost breakdown. Add-ons and monthly fees are separate.
These examples use the same telephony rate for every call segment and the same configuration for both agents. Substitute each segment’s actual rate when the destination or agent configuration differs. With custom telephony, the Retell telephony component is $0; add your carrier’s charges separately.
Rule 1: Minimum duration for dynamic opening messages
When it applies: A call lasts less than 10 seconds and uses a dynamic opening message with AI speaks first. Billing adjustment: Retell bills a minimum duration of 10 seconds.
A dynamic opening message configured for AI speaks first
Rule 2: LLM price scaling for > 4,000 token prompt length
When it applies: The agent uses more than 4,000 LLM tokens in its prompt. Billing adjustment: Retell scales the billed duration in proportion to the token count. The token calculation includes the global prompt, function descriptions, state or node prompts, transcript, tool history and results, and retrieved knowledge base content.Flex mode compiles all node prompts, transitions, and tool descriptions into one LLM context, which can push the token count above 4,000.
- Scaling factor = prompt LLM tokens ÷ 4,000
- Billed duration = original duration × scaling factor, rounded up

