Retell AI Pricing & Cost Breakdown (2026)
Retell AI is priced per minute of call time, and the total is a stack of a few components — mostly the language model you pick. Here's the current breakdown from Retell's pricing page, a worked monthly example, what testing costs, and how to keep the bill down.
✓ Top Rated Plus · ✓ 100% Job Success · ✓ 9+ years experience · ✓ US · UK · EU clients
Quick answer
Retell AI is pay-as-you-go, billed per minute of call time. Retell lists $0.07–$0.31 per minute for voice agents, depending on the language model, voice and phone line you choose. A typical setup — GPT-4.1 ($0.045/min), Retell's voice infrastructure ($0.055/min), a standard voice ($0.015/min) and a Retell phone line ($0.015/min) — comes to about $0.13 per minute. New accounts get $10 of free credit and 20 concurrent calls, and testing is billed at the same rates as live calls.
How Retell AI pricing works
Your per-minute cost is the sum of these parts (Retell's published rates, October 2026):
- Language model — from under $0.01 to $0.16+ per minute depending on the model (the main variable)
- Retell voice infrastructure — $0.055 per minute
- Voice (text-to-speech) — $0.015 per minute for standard voices from Retell, Cartesia, OpenAI and others; $0.04–$0.10 per minute for ElevenLabs voices
- Telephony — $0.015 per minute on a Retell number, or no Retell charge if you bring your own SIP trunk
- Phone numbers — $2 a month each when bought through Retell
- Optional add-ons — for example knowledge base +$0.005/min, advanced denoising +$0.005/min, PII removal +$0.01/min
Want this done for you? I'm a Top Rated Plus AI & Automation expert with 100% job success. Get a quote →
Language model cost comparison
You can swap the model to balance quality and cost — a simple booking agent doesn't need the most expensive one. Per-minute prices from Retell's rate table:
- GPT-4.1 nano — $0.0032/min: very cheap, for simple, high-volume flows
- GPT-5 mini — $0.008/min
- GPT-4.1 mini — $0.0128/min
- Gemini 3.0 Flash — $0.016/min
- GPT-4.1 — $0.045/min: a reliable all-round default
- Claude 5 Sonnet — $0.064/min: stronger reasoning for complex calls
- GPT-5.4 — $0.080/min
What it costs per month: a worked example
An agent handling 500 calls a month at about 3 minutes each uses 1,500 minutes. With GPT-4.1, a standard voice and a Retell number, that's about $0.13 × 1,500 ≈ $195 a month, plus $2 for the phone number — usually far less than staffing the phone for the same hours.
- Cost scales with calls × minutes per call
- Switching to GPT-4.1 mini cuts the same example to roughly $150 a month
- Premium voices add up to $0.085 a minute
- Long prompts can raise the bill: Retell's model rate assumes a prompt within its token allowance, and bills longer ones proportionally
Does testing a Retell AI bot cost money?
Yes. Retell has no free testing sandbox — tests are billed at the same rates as production. The $10 sign-up credit covers early testing of a small agent.
- Web test calls from the dashboard — billed per minute at the voice rate, without the telephony charge
- Text tests in the LLM playground — billed per message at the model's chat rate (fractions of a cent to a few cents per message)
- Simulation tests — bill the agent's model plus the model playing the caller
- AI quality assurance — first 100 minutes free, then $0.10 a minute
- Cheapest approach: test the logic in text first, then make short voice calls to check the voice and timing
Concurrency: how many calls at once
Every account includes 20 concurrent calls. Each extra concurrent call costs $8 a month, and Enterprise plans start at 50+.
How to keep Retell AI costs down
A few smart choices make a big difference.
- Match the model to the task (don't over-pay)
- Keep prompts tight so calls stay efficient
- Use a knowledge base so calls resolve faster
- Route only the right calls to the agent
- Bring your own phone provider over SIP if you already have one