Retell AI Pricing in 2026: The $0.07 to $0.31 Band, Decoded
Retell publishes a pay-as-you-go band of $0.07 to $0.31 per minute, and it publishes the component rates that band is built from. The $0.07 floor is the cheapest possible configuration, not a bundle. A production sales stack lands nearer the middle of the band once you pick a real model, a real voice and a telephony route. Here is the component math for the 2026 procurement cycle.
Last verified June 2026; all rates below re-read from Retell's pricing page on 23 September 2026. Retell pricing changes periodically; always confirm before purchase.
$0.07-$0.31/min
Published voice agent band
$0.055/min
Retell Voice Infra
$0.015/min
Telephony (Twilio/Telnyx via Retell)
Custom
Enterprise
§The Component Rates Retell Actually Publishes
Retell prices four separate lines and then quotes a band across all of them. It is not a bundle. Picking a cheap model and a cheap voice puts you at the bottom of the published band; picking a frontier model and ElevenLabs puts you at or above the top.
| Line | Published rate | Notes |
|---|---|---|
| Retell Voice Infra | $0.055/min | Retell's own layer, charged on every minute |
| TTS: Retell Platform, Minimax, Fish, Cartesia, OpenAI, Inworld | $0.015/min | All six priced identically |
| TTS: ElevenLabs | $0.040/min | The only voice provider priced above the $0.015 tier |
| LLM: cheapest published | $0.0016/min | GPT 5 nano; GPT 5.6 Luna and GPT 5.4 nano both $0.0064 |
| LLM: recommended tier | $0.064-$0.16/min | GPT 5.6 Terra and Claude 5 Sonnet $0.064; GPT 5.4 $0.080; GPT 5.5 $0.16 |
| LLM: most expensive published | $0.32/min | GPT 6 Astra standard tier; $0.64 on the fast tier |
| Telephony | $0.015/min | Retell's Twilio or Telnyx routing; no charge for SIP trunking or custom telephony |
Published add-on meter
- + Knowledge base: $0.005 per minute
- + Advanced denoising: $0.005 per minute
- + Safety guardrails: $0.005 per minute
- + PII removal: $0.01 per minute
- + AI quality assurance: first 100 minutes free, then $0.10 per minute
- + Batch call: $0.005 per dial
- + Branded call: $0.10 per outbound call
- + SMS: $0.01 per message
What pay-as-you-go includes at no extra charge
- + $10 in free credits on signup
- + 20 free concurrent calls
- + Full platform access and pre-built agent templates
- + Call analytics and transcripts, simulation testing, webhooks and API access
- + Community and email support
- + No charge for SIP trunking or custom telephony if you bring your own carrier
§All-In Cost: Three Reference Sales Stacks
These three stacks are built only from rates Retell publishes, added up. They land at $0.076, $0.149 and $0.270 per minute, which sits inside Retell's own published $0.07 to $0.31 band and shows where within it a real configuration falls.
| Component | Budget | Mid | Premium |
|---|---|---|---|
| Retell Voice Infra | $0.055 | $0.055 | $0.055 |
| TTS | $0.015 (Retell Platform voice) | $0.015 (Retell Platform voice) | $0.040 (ElevenLabs) |
| LLM | $0.0064 (GPT 5.4 nano) | $0.064 (GPT 5.6 Terra) | $0.16 (GPT 5.5) |
| Telephony | $0 (own SIP trunk) | $0.015 (Retell Twilio/Telnyx) | $0.015 (Retell Twilio/Telnyx) |
| Per-minute total | $0.076 | $0.149 | $0.270 |
| Per 5-min call | $0.38 | $0.75 | $1.35 |
The Retell-vs-Vapi math, corrected: Vapi's own pricing calculator estimates a 1,000-minute month at $82 to $129, which is $0.082 to $0.129 per minute all-in on Vapi's free telephony, covering its $0.05 hosting fee, Deepgram transcription at $0.0095 to $0.0099, an OpenAI model at $0.0077 to $0.0452 and an ElevenLabs voice at $0.0146 to $0.0238. Retell's mid stack at $0.149 per minute is above the whole of that range. Earlier versions of this page put Vapi at $0.25 to $0.33 per minute all-in and concluded Retell was about 21 percent cheaper; on both vendors' published numbers that conclusion is wrong and the direction reverses. The trade is real but it is not price: Retell bills one invoice with a published add-on menu, while Vapi passes model costs through at cost and sells support separately.
§Monthly Volume Math for Sales Teams
| Scenario | Monthly min | Retell mid ($0.149) | Vapi published ($0.082-$0.129) | Cheaper option |
|---|---|---|---|---|
| Pilot: 3K min | 3,000 | $447 | $245-$387 | Vapi by $60-$202 |
| Mid: 15K min | 15,000 | $2,235 | $1,227-$1,934 | Vapi by $301-$1,008 |
| High: 50K min | 50,000 | $7,450 | $4,090-$6,445 | Vapi by $1,005-$3,360 |
| Scale: 250K min (both quote-only at this size) | 250,000 | $37,250 at list | $20,450-$32,225 at list | Neither publishes a volume tier |
On published list rates, Vapi is 13 to 45 percent cheaper per minute than a mid Retell stack, and because both price linearly the gap holds at every volume shown. Neither vendor publishes a volume tier, so the 250,000-minute row is list arithmetic rather than a quote; both say enterprise pricing is custom. Retell's counter-arguments are a single invoice, a published add-on menu and free telephony if you bring your own SIP trunk, which is a genuine saving at $0.015 a minute. Teams that need fine-grained control over each component should evaluate Vapi.
§Latency Reality Check
Retell publishes a 250 to 400ms first-token-to-first-audio latency for the standard inbound configuration. This is competitive with the best Vapi configurations and reflects Retell's tight integration of Deepgram streaming STT with their orchestration layer.
The latency band widens for outbound calls because the carrier setup time (Twilio's outbound dial sequence) adds 1 to 3 seconds before the AI's first turn. For inbound qualification this is invisible; for outbound cold call openings, the prospect may hear silence at the start of the call and hang up before the AI speaks. Mitigations: pre-recorded greeting, ringback before answer detection, optimised SIP trunk routing.
Premium LLM choices (GPT-4o full, Claude Sonnet 3.5/3.6) add 200 to 500ms to the model-reasoning portion of latency. For complex objection handling this is worth the latency hit; for simple appointment-booking the budget LLM at sub-300ms total is the better experience trade-off.
§Retell vs Bland vs Vapi: Pricing Summary
| Platform | Headline rate | All-in typical | Monthly minimum | Strength |
|---|---|---|---|---|
| Retell AI | $0.07-$0.31/min published band | $0.076-$0.270/min on the stacks above | None | One invoice; published add-on menu; free BYO SIP |
| Vapi | $0.05/min hosting | $0.082-$0.129/min (Vapi's own calculator) | None on usage; Core $29/mo, Pro $999/mo minimum | Models pass through at cost; free Vapi telephony |
| Bland AI | $0.14/min Start, $0.12/min Build | $0.12-$0.14/min plus $0 or $299/mo platform fee | Plan-based | Built-in dialer + sequencing; no model pass-throughs |
| Synthflow | Not published | Not published | Enterprise contracts start at $30,000/yr | No-code friendly; enterprise-only pricing |
On published list rates, Vapi is the cheaper per-minute option on most realistic sales configurations and Bland is the simplest to model because it quotes one all-in number per tier with no model pass-throughs. Retell wins where operational simplicity and a published add-on menu matter more than the per-minute delta. The full head-to-head is at Vapi vs Retell.