I’m evaluating the economics of deploying AI voice agents on Retell and would appreciate some guidance on pricing benchmarks.
What is the typical average cost per minute for voice agents built on Retell? I’m particularly interested in understanding the total cost, including telephony, speech-to-text, LLM, and text-to-speech components.
If possible, could you share:
Typical per-minute costs for customer support voice agents
Cost ranges for simple vs. more complex agents with tool calling and integrations
Any real-world benchmarks or examples from production deployments
I’m trying to estimate expected call costs and monthly spend for a support use case.
There’s heaps of huge if/buts/maybes in this scenario.
It depends on your telephony charges, ie if your using twilio they will have their own SIP trunking rate you need to account for. Unless you’re in the USA these will have to be managed external to retell (as far as I am aware)
It depends on the compelxity of the prompt/agent - if you have a single agent prompt but are using 10,000 token Agent prompt then your cost will be higher than someone who has a much shorter agent prompt.
Middleware/API services - are you connecting your agent externally, you will need to account for these? (can be a fraction of a cent per api call depending on your setup)
To give you a super rough Idea, I have a bunch of clients and so far YTD we have found the average to be around $0.17 US per minute to be the average. This is JUST the retellai cost.
This accounts for a mix of both higher complexity and lower complexity agents.
Hello @sabari.selvaraj Retell is usage-based and the per-minute cost is the sum of the components you pick, with no flat per-agent fee on the base plan.
Voice Engine (always applied): $0.055/min
Text-to-Speech: $0.015/min for Retell Platform / MiniMax / Cartesia / Fish / OpenAI voices; $0.040/min for ElevenLabs
LLM: varies by model and standard-vs-fast tier — the full current model list and rates live on the pricing page below (we refresh it as model lineups change, so I’d rather point you at the live numbers than quote a snapshot)
Telephony: managed Twilio numbers are billed at our pass-through rate on the pricing page; BYOC (bring your own Twilio/Telnyx/SIP) is $0/min from Retell and you pay your carrier directly
Optional add-ons (Knowledge Base, Background denoising, PII redaction, AI QA, etc.) are listed on the same page with their per-minute deltas
The page also has an interactive calculator where you can pick a voice + LLM + add-ons and see the resulting $/min in real time, which is the cleanest way to get “simple agent” vs “complex agent” ranges for your stack.
@sabari.selvaraj Let me share my experience. Using it if you use a simple agent without complex features, it might be around 11 - 15 cents/minutes, but if you have more complex it can go beyond that. My highest would be around 16-23 cents/minutes. It still can go up based on the function and purpose of the agent
@sabari.selvaraj For a high quality agent (11Labs V3 Voices) with production level knowledge (IE 6k plus tokens) you’re looking around $0.35-$0.40 per minute.
one of the beautiful things about Retell is the ability to customize your costs based on your use case. For my clients, typical would be $0.12-$0.14 per minute.