Retell AI
Retell AI pricing breaks a voice minute into named parts and publishes a rate for each.
Updated on:
Retell AI pricing: the four parts of a voice minute
Retell AI pricing breaks a voice minute into named parts and publishes a rate for each. You pay $0.055 a minute for Retell's voice engine, plus a rate for the model you picked, plus a rate for the voice, plus telephony. Three of those four you can change or remove. The headline $0.07 to $0.31 range on the pricing page is the part that doesn't survive arithmetic.
Key takeaways
Retell charges $0.055 a minute for voice infrastructure and prices the LLM, voice and telephony as separate invoice lines.
Bring your own LLM and the LLM line goes to zero. Bring your own SIP trunk and telephony goes to zero.
Concurrency costs $8 per concurrent call per month beyond the free 20, and burst calls add $0.10 a minute to the whole call.
Prompts over 4,000 tokens scale billed duration: a 60 second call on a 4,800 token prompt bills as 72 seconds.
Retell AI pricing in 2026
Tier | Price | Included | Metered limit
|
|---|---|---|---|
Pay as you go | $0 platform fee | Full platform, API, analytics | 20 concurrent calls, $10 credit |
Enterprise | Custom | SSO, RBAC, custom MSA/DPA/BAA | Concurrency from 50+, volume pricing |
Component | Rate | ||
--- | --- | ||
Retell voice infra | $0.055/min | ||
Voices: Platform, Minimax, Fish, Cartesia, OpenAI, Inworld | $0.015/min | ||
Voices: ElevenLabs | $0.040/min | ||
LLM | $0.0016/min (GPT 5 nano) to $0.64/min (GPT 6 Astra, fast tier) | ||
LLM, recommended models | $0.064/min (GPT 5.6 Terra, Claude 5 Sonnet) | ||
Telephony, Retell Twilio US | $0.015/min | ||
Custom LLM or custom telephony | No charge |
What Retell AI actually meters
Connected call seconds, split across four components and totalled by the minute.
Four line items, and a fifth that never appears. Speech to text has no published rate of its own. Retell folds transcription and turn-taking into the $0.055 voice infra line and picks the vendor itself, failing over between Deepgram and Azure mid-call. You can't swap the transcriber or see what it costs.
The other three you control. The LLM line carries a per-minute rate for each of 26 models, several with a fast tier at double the standard rate. Point Retell at your own model and the line reads "No charge". Text to speech is $0.015 a minute across six voice providers and $0.040 for ElevenLabs. Telephony runs $0.015 a minute on a Retell-managed Twilio US number, up to $0.80 for the Philippines, and nothing on your own trunk.
Two adjustments change billed duration rather than the rate. A dynamic opening message carries a 10 second minimum, and prompts above 4,000 tokens scale duration by tokens divided by 4,000, rounded up. Silence bills, because the transcriber keeps listening. Unconnected calls don't.
How credits work
Newer workspaces run on prepaid credits; older ones bill monthly in arrears.
Credit accounts deduct usage from a balance in real time. Credits never expire and are non-refundable once bought, an unusual pairing: most vendors trade one for the other. Balances are scoped per workspace, so a card added in one doesn't carry to the next.
Subscription items sit outside the credit balance and bill to the card at month end: numbers at $2, verified numbers at $10, SMS at $20, knowledge bases beyond the first ten at $8, and concurrency at $8 a slot, charged on purchase and prorated by day since February 2026.
Free allowances: $10 of signup credit once per email address, 20 concurrent calls, ten knowledge bases, the first 100 AI QA minutes, and 30 Conductor messages per user per day.
What happens when you hit the limit
Two limits, two different failures.
Run out of credits and calls stop. Retell blocks new calls at a zero balance until auto recharge fires on a threshold and target you set, so the refill is a fixed top-up, not metered overage. Legacy monthly accounts don't block; they bill in arrears.
Run out of concurrency and it depends on direction. Outbound calls are rejected. Inbound calls wait about 40 seconds, then transfer to a fallback number or end with concurrency_limit_reached. Turn on burst and they go through, capped at the lower of 3x your limit or your limit plus 300, with $0.10 a minute added to the entire call, not just the part above the line.
How Retell AI pricing has changed across all these years
Date | Milestone | Source
|
|---|---|---|
22 Sep 2026 | Inworld joins the $0.015 voice tier. Component rates and the $0.07 to $0.31 range unchanged | Vendor |
2 Jul 2026 | Free LLM token allowance raised from 3,500 to 4,000 | Vendor |
10 Feb 2026 | TTS and voice engine split into separate invoice lines. Concurrency billed upfront, prorated daily | Vendor |
16 May 2025 | Token scaling above 3,500 prompt tokens and a 10 second minimum on dynamic openers, from 1 June | Vendor |
31 Dec 2024 | GPT-4o Realtime cut from $1.50 to $0.50/min. Telephony raised from $0.01 to $0.015/min | Vendor |
19 Apr 2024 | Pricing made modular: voice engine, LLM and telephony split into separate rates, custom LLM free | Vendor |
4 Mar 2024 | Discounted enterprise tiered pricing introduced | Vendor |
6 Feb 2024 | New $0.10/min tier using OpenAI TTS | Vendor |
Customer
Sentiment Highlights
"I do wish Retell's pricing was cheaper, though; $6/hr is pretty much the cost of a call center employee in India"
reissbaker, Hacker News, February 2024
"The pricing can also become expensive when scaling or doing many test calls during development."
Verified user in computer software, G2, June 2026
Explore other providers
Replicate
Infrastructure Platform
Replicate pricing charges per second of compute, and the per-second rate depends on which GPU your model runs on.
Pinecone
Data Platform
Pinecone pricing meters four things on a serverless index: read units, write units, gigabytes stored and gigabytes returned.
Perplexity API
API
Perplexity API pricing charges you twice for the same call.
How much does Retell AI cost per minute?
Does Retell AI charge a platform fee?
How much does Retell AI concurrency cost?
Do Retell AI credits expire?
