Voice AI Pricing 2026: Vapi vs Bland vs Retell

Verified as of 2026-06-16 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 2026 product. Their headline numbers are not comparable: one quotes a thin platform fee, one quotes a single bundled rate, and one quotes a stack of line items. This article normalizes all three to an all-in cost per minute, prices a concrete 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 $0.11 (Bland) to $0.07 (Retell) as if they measured the same thing. They do not. Vapi's $0.05 covers orchestration only; the models that actually run the call are extra^[1]. Bland's rate already contains the models^[2]. Retell's lowest figure is its voice infrastructure line before TTS, LLM, and telephony are added^[3]. To compare honestly, build each one up to a full, working call and read the total.

2. Vapi: a thin platform fee over at-cost passthrough

Vapi charges a $0.05 per-minute platform fee for hosting a call^[1]. Speech-to-text, the language model, and text-to-speech are billed at cost from the providers, or $0 if you bring your own API key, so they land on top of the $0.05 rather than inside it^[1]. The Build plan includes 60+ minutes and 10 concurrent calls, with extra lines at $10 per line per month; the Scale plan moves to a fixed platform fee plus committed, volume-based per-minute pricing^[1].

Because the models pass through at cost, the true per-minute number depends on which STT, LLM, and TTS you wire up. Stacking a transcription engine, an LLM, a voice model, and a telephony carrier onto the $0.05 platform fee puts a real Vapi minute around $0.15 to $0.31^[1]. That is the figure to budget against, not the $0.05 headline. The upside is full control: swap in a cheaper LLM and the all-in drops; bringing your own keys zeroes the model line and leaves you paying mostly the platform fee plus telephony.

3. Bland: one bundled rate, no passthroughs

Bland takes the opposite approach. Its per-minute Talk rate bundles the LLM, speech-to-text, text-to-speech, and telephony into a single number, stated on the pricing page as having no token charges and no provider passthroughs^[2]. The Talk rate is $0.14 per minute on the free Start tier, $0.12 on Build ($299 per month), and $0.11 on Scale ($499 per month); call transfers bill separately at $0.04 to $0.05 per minute^[2].

Tier limits scale with the monthly fee: Start allows 10 concurrent calls and 100 calls per day, Build allows 50 concurrent and 2,000 per day, and Scale allows 100 concurrent and 5,000 per day^[2]. New accounts get two free credits plus a free inbound number, with no platform fee and no card required^[2]. The trade is predictability for control: you get one number to plan against, but you do not pick the underlying STT, LLM, or TTS the way you do on Vapi or Retell.

4. Retell: a stack of separately billed line items

Retell prices each layer of the call as its own line. As of June 2026 the pay-as-you-go plan lists Retell voice infrastructure at $0.055 per minute, platform text-to-speech at $0.015 per minute, the LLM at $0.003 to $0.16 per minute depending on the model, and US telephony at $0.015 per minute^[3]. The stated range is $0.07 to $0.31 per minute; building up a representative configuration from those line items (GPT-5.1 plus Retell TTS plus US telephony) lands near $0.127 per minute^[3].

The LLM line is what moves the total: Gemini 2.5 Flash Lite is listed at about $0.006 per minute while GPT-5.5 is $0.16 per minute, so the same agent can sit at either end of the range purely on model choice^[3]. Add-ons stack further: a knowledge base or advanced denoising adds $0.005 per minute each, and PII removal adds $0.01 per minute^[3]. New accounts get $10 in free credits and 20 free concurrent calls, with extra concurrency at $8 per slot per month^[3].

5. All-in per-minute, normalized

All figures are pay-as-you-go all-in cost per call minute, verified June 17, 2026. The Vapi and Retell rows depend on the models you select; the Bland row is fixed by tier.

Platform	Billing model	Headline	Real all-in / min
Vapi	Platform fee + at-cost STT/LLM/TTS[1]	$0.05/min fee[1]	~$0.15-$0.31[1]
Bland (Start, free)	Bundled, no passthroughs[2]	$0.14/min[2]	$0.14[2]
Bland (Build, $299/mo)	Bundled, no passthroughs[2]	$0.12/min[2]	$0.12[2]
Bland (Scale, $499/mo)	Bundled, no passthroughs[2]	$0.11/min[2]	$0.11[2]
Retell (low LLM)	Unbundled line items[3]	$0.07/min[3]	~$0.09-$0.13[3]
Retell (high LLM)	Unbundled line items[3]	up to $0.31/min[3]	~$0.25-$0.31[3]

Read across, not down. Bland's bundled $0.11 to $0.14 is the most predictable; Vapi and Retell can dip below it with a cheap LLM and your own keys, or climb above it with a flagship model and add-ons. The headline column shows why the comparison fails at a glance: $0.05, $0.11, and $0.07 are three different layers of the call.

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

Scenario: an outbound or support agent handling 20,000 call minutes per month, a mid-volume solo or small-team load. Using each platform's verified rates, here is the rough monthly spend. Vapi and Retell assume a mid-cost LLM around the $0.13 to $0.20/min all-in band; Bland uses its fixed tier rate.

• Vapi: $0.05/min platform fee = $1,000, plus at-cost STT, LLM, TTS, and telephony. At a mid-stack ~$0.13/min on top, the all-in lands near $0.18/min, or roughly $3,600/month^[1]. Bring your own keys and the model line shrinks toward the platform fee plus telephony.
• Bland (Build): $0.12/min × 20,000 = $2,400 in talk time, plus the $299/month platform fee = about $2,699/month, fully bundled^[2]. On Scale ($0.11/min + $499) it is about $2,699/month at this volume, so Build and Scale roughly converge near 20,000 minutes.
• Retell: at a ~$0.127/min built-up config (GPT-5.1 + Retell TTS + US telephony), 20,000 minutes is about $2,540/month; a cheap-LLM setup drops toward $0.09/min (~$1,800), a flagship-LLM setup climbs toward $0.31/min (~$6,200)^[3].

At this volume the bundled and unbundled approaches land close: Bland Build (~$2,700) and Retell's mid config (~$2,540) are within a few hundred dollars, while Vapi's all-in depends entirely on whether you bring your own model keys. The decisive variable is not the platform; it is the LLM and whether you self-supply keys. Fold the result into your full monthly budget with the AI stack cost calculator.

7. Which to pick

1. Pick Bland if you want one predictable number and do not need to choose the underlying models. The bundled $0.11 to $0.14/min removes passthrough surprises, and the free Start tier plus a free inbound number lets you test before paying a platform fee.
2. Pick Vapi if you want maximum control and plan to bring your own provider keys. The $0.05 platform fee plus at-cost models is the cheapest floor when you self-supply LLM and TTS keys, but budget for the real $0.15 to $0.31/min if you let Vapi pass models through at cost.
3. Pick Retell if you want itemized control without managing keys, and you want to dial cost with model choice. A cheap LLM lands near $0.09/min; a flagship pushes past $0.30. Read the line items before committing.
4. Either way, re-verify rates and price your own LLM choice before signing up. The model line, not the platform fee, decides the bill.

Re-verify each pricing page before committing; voice-AI rates and model line items change often. For the underlying model costs that drive every voice agent's bill, see the cheapest LLM API ranking and the Deepgram vs AssemblyAI speech-to-text comparison. If voice is one channel inside a wider support operation, the best AI customer support tools roundup covers the per-ticket side.

All rate figures verified against official pricing pages as of 2026-06-17.

Frequently asked questions

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

It depends on what you count. Bland quotes the lowest single all-in number: $0.11 to $0.14 per minute bundled, with no separate provider charges. Vapi's headline is $0.05 per minute, but that is only the platform fee; once you add at-cost speech-to-text, an LLM, and text-to-speech, a real call lands near $0.15 to $0.31. Retell sits between, with its components stacking to about $0.07 to $0.31 depending on the LLM you choose. For a no-tuning fixed rate, Bland is cheapest; for a tuned, model-choice setup, Vapi or Retell can go lower or higher than Bland.

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

Because $0.05 per minute is the Vapi platform fee alone. Speech-to-text, the language model, and text-to-speech are billed at cost from the providers, or $0 if you bring your own API key, so they are added on top of the platform fee. A typical call stacks transcription, an LLM, a voice engine, and telephony onto the $0.05, which pushes a real minute to roughly $0.15 to $0.31 depending on the models you pick. The $0.05 is the orchestration layer, not the whole call.

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

No. Bland bundles the language model, speech-to-text, text-to-speech, and telephony into one per-minute Talk rate with no token charges and no provider passthroughs, per its pricing page. The rate is $0.14 per minute on the free Start tier, $0.12 on Build ($299 per month), and $0.11 on Scale ($499 per month). Transfers are billed at a lower separate rate ($0.04 to $0.05 per minute). The trade-off is a single predictable number against less control over which underlying models you run.