Voice AI Pricing 2026: Vapi vs Bland vs Retell

Verified on 2026-10-11 against the official vendor pricing pages cited below.

Vapi, Bland and Retell are the three voice-agent platforms a builder weighs when adding phone calls, voice support or outbound dialing to a product in 2026. Their headline numbers are not comparable: one quotes a hosting fee, one quotes a bundled AI rate, and one quotes a stack of line items. This article normalizes all three to an all-in cost per minute, prices a 20,000-minute month, and lays the verified rates side by side.

1. The headline fee is not the price

The mistake is comparing $0.05 (Vapi) to $0.12 (Bland) to $0.07 (Retell) as if they measured the same thing. Vapi's $0.05 covers hosting only; the models that run the call are extra^[1]. Bland's rate already contains the models but not telephony^[2]. Retell's $0.07 floor is a full stack with its cheapest options^[3]. To compare honestly, build each one up to a working call and read the total.

2. Vapi: platform fee over at-cost models

Vapi charges $0.05 a minute in hosting and passes speech-to-text, the LLM and text-to-speech through at cost^[1]. Its pricing estimator lists Deepgram transcription at $0.0095 to $0.0099 a minute, OpenAI models at $0.0077 to $0.0452 and ElevenLabs voices at $0.0146 to $0.0238; Vapi's own telephony is free, while Twilio adds $0.008 a minute inbound and $0.014 outbound^[1]. For 1,000 minutes the estimator totals $82 to $129^[1].

Plans are now success packages layered on usage. With no package you start free with $5 in credits, 4 concurrent calls and one phone number. Core is $29 a month with 10 concurrent calls; Pro is 10% of the hosting fee with a $999 monthly minimum and 30 concurrent calls; extra concurrency is $10 per line a month^[1]. The upside is control: a cheaper LLM or voice lowers the all-in directly.

3. Bland: one bundled AI rate

Bland bundles the LLM, real-time transcription and premium voices into one per-minute rate with no token charges^[2]. Start costs $0.14 a minute with no platform fee, 10 concurrent calls and 100 calls a day; Build costs $0.12 a minute plus a $299 monthly platform fee, with 50 concurrent calls and 2,000 calls a day. Transfer time is $0.05 or $0.04 a minute, and Enterprise is contracted to volume^[2].

Telephony is billed separately: bring your own Twilio account or SIP trunk, or use Bland's built-in Twilio at pass-through cost^[2]. New Start accounts get two credits and an inbound number valued at $15 a month, with no card required^[2]. The trade is predictability for control: one number to plan against, without choosing the underlying models.

4. Retell: separately billed line items

Retell prices each layer as its own line: voice infrastructure $0.055 a minute, platform text-to-speech $0.015 a minute (ElevenLabs voices $0.04 to $0.10), the LLM from $0.0016 a minute for GPT 5 nano to $0.16 for GPT 5.5 on the standard tier, and US telephony $0.015 a minute^[3]. A GPT 5.1 configuration ($0.04) with Retell voices and US telephony totals $0.125 a minute^[3].

Add-ons stack: a knowledge base or advanced denoising adds $0.005 a minute each and PII removal $0.01. New accounts get $10 in credits and 20 free concurrent calls, with extra concurrency at $8 a month each^[3].

5. All-in per-minute, normalized

Platform	Billing model	Headline	All-in / min
Vapi (estimator defaults)	Hosting + at-cost models[1]	$0.05 hosting[1]	~$0.08-$0.13 (Vapi telephony)[1]
Bland Start	Bundled AI rate[2]	$0.14[2]	$0.14 + telephony
Bland Build ($299/mo)	Bundled AI rate[2]	$0.12[2]	$0.12 + telephony
Retell (GPT 5 mini)	Line items[3]	from $0.07[3]	$0.093
Retell (GPT 5.1)	Line items[3]	-	$0.125
Retell (GPT 5.5)	Line items[3]	up to $0.31[3]	$0.245

The Retell rows add $0.055 infrastructure, $0.015 voice and $0.015 US telephony to the LLM line ($0.008 for GPT 5 mini, $0.04 for GPT 5.1, $0.16 for GPT 5.5)^[3]. Vapi with its default estimator stack is now the lowest all-in of the three, Retell overlaps it with a small model, and Bland's bundled rate is the most predictable but excludes telephony.

6. A 20,000-minute month, priced three ways

Scenario: a support or outbound agent handling 20,000 call minutes a month, using each platform's verified rates:

• Vapi: $0.05 × 20,000 = $1,000 hosting, plus models at cost. On the estimator's range ($0.082 to $0.129 a minute all-in) that is about $1,640 to $2,580 a month on Vapi telephony, plus a success package if you need more than 4 concurrent calls^[1].
• Bland Build: $0.12 × 20,000 = $2,400 plus the $299 platform fee = $2,699 a month before telephony^[2]. On Start ($0.14, no platform fee) it would be $2,800, but the 100-calls-a-day cap rules Start out at this volume.
• Retell: $0.125 × 20,000 = $2,500 on the GPT 5.1 setup; $1,860 with GPT 5 mini; $4,900 with GPT 5.5^[3].

At this volume the three land within about $1,000 of each other on mid-range models; the LLM choice moves the bill more than the platform does. Fold the result into your monthly budget with the AI stack cost calculator.

7. Which to pick

1. Pick Bland if you want one predictable AI rate and do not need to choose models. Budget telephony on top.
2. Pick Vapi if you want the lowest all-in with control over models and free built-in telephony, and you are comfortable adding a success package as concurrency grows.
3. Pick Retell if you want itemized pricing without managing provider keys and 20 free concurrent calls from day one.
4. Either way, price your own LLM and voice choice before signing up; those lines decide the bill.

Re-verify each pricing page before committing; all three changed their plans in 2026. For the model costs behind every voice agent, see the cheapest LLM API ranking and the Deepgram vs AssemblyAI speech-to-text comparison. If voice is one channel in a wider support operation, see the best AI customer support tools.

All rate figures verified against official pricing pages on 2026-10-11.

Frequently asked questions

Is Vapi, Bland or Retell cheapest per minute in 2026?

It depends on the models you choose. Vapi's own estimator, at $0.05 a minute hosting plus Deepgram, OpenAI and ElevenLabs at cost, lands at about $0.08 to $0.13 a minute on Vapi telephony. Bland bundles the LLM, speech-to-text and voice into $0.14 a minute on Start or $0.12 on Build ($299 a month), with telephony billed separately. Retell itemizes to $0.07 to $0.31 a minute; a GPT 5.1 setup with Retell voices and US telephony is about $0.125. Verified on each pricing page on 2026-10-11.

Why is Vapi's real cost higher than its $0.05 per minute?

Because $0.05 a minute is Vapi's hosting fee alone. Speech-to-text, the language model and text-to-speech are passed through at cost from the providers. Vapi's own estimator lists Deepgram at about $0.0095 to $0.0099 a minute, OpenAI models at $0.0077 to $0.0452 and ElevenLabs voices at $0.0146 to $0.0238, which puts a typical minute at $0.08 to $0.13 before any third-party telephony; Vapi's own telephony is free.

Does Bland AI charge separately for the LLM, voice and telephony?

Bland's per-minute rate covers the language model, speech-to-text and text-to-speech with no token charges or model pass-throughs. Telephony is now billed separately: you can use your own carrier or SIP trunk, or Bland's built-in Twilio at pass-through cost. The rate is $0.14 a minute on Start and $0.12 on Build, which has a $299 monthly platform fee; transfer time is $0.05 or $0.04 a minute.

What does Retell AI cost per minute?

Retell's pay-as-you-go range is $0.07 to $0.31 a minute. The components are voice infrastructure at $0.055 a minute, text-to-speech from $0.015 (platform voices) to $0.10 (some ElevenLabs voices), the LLM from $0.0016 a minute for GPT 5 nano to $0.16 for GPT 5.5 on the standard tier, and US telephony at $0.015 a minute. New accounts get $10 in credits and 20 free concurrent calls.