OfficeBooks
Voice AI agent · Comparison

Vapi Alternatives Ranked by All-In Cost Per Minute

Affiliate disclosure: some links here are affiliate links and we may earn a commission at no cost to you. The order below follows the stated criterion and nothing else; no vendor paid for a position or saw this page before it went up.

Vapi’s headline is $0.05 a minute, and that is not what a minute costs. It is Vapi’s hosting fee. The same page states models pass through at cost, so Deepgram transcription at $0.0095-$0.0099/min, OpenAI at $0.0077-$0.0452/min and ElevenLabs voice at $0.0146-$0.0238/min stack on top, then Twilio carrier minutes at $0.008 inbound or $0.014 outbound. The only button a visitor can press without sales sits on the $0 prepaid tier with its $5 of credits: the $29/month Core plan’s call to action reads Contact Sales.

So the real question behind Vapi alternatives is not who quotes a smaller number, but whose published pages let you total a minute at all. This ranking measures the all-in cost once telephony and model charges are counted, and whether you can sign up and ship without a sales call. The short answer: Amazon Connect publishes every layer on its own site and lists an AI voice agent minute at about $0.046 before telephony, with US carrier rates alongside it. Retell is the closest voice-agent-native equivalent, the one platform where the carrier line is both priced and removable.

What this ranking measures

This page ranks ten platforms on one thing: how much of a voice minute you can price from the vendor’s own published pages — the agent or hosting fee, speech-to-text, the model, the voice, the carrier, and any platform subscription underneath — and whether you can sign up and ship without a sales call. A vendor that leaves a layer unpriced loses position for it, because an unpriced layer still lands on your invoice. What this deliberately does not measure: voice quality, real-world latency, transcription accuracy, tooling, support or integrations. Nobody here installed, subscribed to or ran any of these platforms. Every figure below was read from published pricing pages on 2026-09-17, and where a vendor publishes nothing, that absence is reported as the finding.

The Vapi pricing calculator set to 1,000 minutes a month, itemising a $50 hosting fee, $10 for the Deepgram transcriber, $8 to $45 for the OpenAI model, $15 to $24 for the ElevenLabs voice and $0 transport.
This is the whole argument of the ranking in one screenshot. The advertised per-minute rate is the hosting line; the transcriber, the model and the voice are billed on top. Screenshot of vapi.ai, captured September 17, 2026. Not a mock-up.

The short answer

Top pickAmazon Connect (Connect Customer)

Every layer of the minute carries a published AWS rate: $0.038 per voice min for the channel, $0.0080 per voice minute for the AI agent — about $0.046 before telephony — plus US carrier at $0.0022 inbound and $0.0048 outbound and $0.0300 per day for a number. Entry is an ordinary AWS account with no minimum and no contract.

Runner-upRetell AI

Retell prices every layer too — $0.055/minute Voice Infra, the model at its own rate, $0.015/minute for platform voices, $0.015/min telephony — and states there is no charge for SIP trunking or custom telephony, so the carrier line can go to zero. Self-serve with $10 in credits, no platform subscription and 20 concurrent calls included.

Cheapest that qualifiesElevenLabs (ElevenAgents) Free

$0 per month with 15 minutes of calls and 4 concurrent calls, as a standing plan rather than a countdown trial, and additional minutes at $0.080. If 15 minutes runs short, Starter is $6 per month for 75 minutes and Creator is $22 per month ongoing with the first month at $11.

All 10 compared

Ranked against the criterion above. Prices are the vendor’s own wording, read on September 17, 2026; where a first term and a renewal differ, both are shown.
#ProductBest forPrice fromWhat decides its place
1 Amazon Connect (Connect Customer) Lowest fully published all-in minute $0.038 per voice min (standard telephony rates apply) $0.0080 AI agent minute on top of the $0.038 voice channel rate Visit
2 Retell AI Builders who want every cost line on one page $0.07-$0.31 / min for AI Voice Agents $0.055/min infra, $0.015/min carrier — or $0 on your own SIP trunk Visit
3 Deepgram Voice Agent API One bundled rate for speech, model and voice $0.075/min (Pay As You Go) / $0.068/min (Growth) $0.075/min bundles STT, model and voice; no telephony rate published Visit
4 Twilio ConversationRelay Keeping the agent and the carrier on one bill $0.07 / minute $0.07/min agent, $0.0140/min carrier, model billed by you Visit
5 Google Cloud Conversational Agents Per-second billing inside Google Cloud $0.001 / 1 second $0.001 per second, and no telephony rate for the current product Visit
6 ElevenLabs (ElevenAgents) A few hundred minutes a month, self-serve Free $0 per month; Starter $6 per month; Creator $22 per month, first month $11 15 minutes free monthly; extra minutes $0.080, burst $0.160 Visit
7 Bland AI One flat rate with no model shopping $0.14 /min (Start) — $0 platform fee. No card required. $0.14/min bundles the model; telephony priced nowhere Visit
8 RingCentral AI Receptionist A finished receptionist, not a platform Starts at $49 /month 100 minutes included, then $0.50/min in 30-second increments Visit
9 Aircall AI Voice Agents Teams already paying for Aircall seats $0.37/min for up to 2,500 minutes, $0.29/min over 2,500 minutes $0.37/min self-serve, plus a $90/month three-licence floor Visit
10 Synthflow Enterprise buyers with a sales cycle Enterprise contracts start at $30,000 annually. $30,000 annually is the only price on the page Visit

1. Amazon Connect (Connect Customer)

Best for: Lowest fully published all-in minute · From: $0.038 per voice min (standard telephony rates apply)

Why it sits at 1 $0.0080 AI agent minute on top of the $0.038 voice channel rate

Amazon Connect Customer is AWS’s contact-centre product, and the AI voice agent is a line item inside it rather than a standalone platform: $0.0080 per voice minute for Connect AI agents, charged on top of the $0.038 per voice min channel rate, both billed per second with a 10-second minimum. The channel rate bundles Amazon Polly generative TTS, Amazon Nova Sonic and Amazon Lex requests that originate in the instance.

It leads because AWS publishes every layer on its own pages: the agent charge sits on top of the channel, giving a listed AI voice minute of about $0.046 before telephony, and US carrier minutes are published at $0.0022 inbound and $0.0048 outbound in us-west-2 with a DID at $0.0300 per day. AWS never adds them up for you, and nothing is free — the Free Tier covers 1,000 profiles per month and 500 chat analytics messages in the first 12 months, with no free voice minutes published.

The Amazon Connect Customer pricing page, headed No minimums, no long-term contracts, pay for what you need, with buttons to get started for free or request a pricing quote.
AWS meters everything and commits to nothing monthly, which suits volume and punishes a business that just wants a predictable bill. Screenshot of aws.amazon.com, captured September 17, 2026. Not a mock-up.
GOOD
  • Agent, channel and US carrier all carry published rates on AWS’s own pages, with no minimum and no contract.
BAD
  • No free voice minutes exist, no latency figure of any kind is published, and outbound campaign dialling costs $0.045 per minute instead of $0.038.

2. Retell AI

Best for: Builders who want every cost line on one page · From: $0.07-$0.31 / min for AI Voice Agents

Why it sits at 2 $0.055/min infra, $0.015/min carrier — or $0 on your own SIP trunk

Retell shows its working. The $0.07-$0.31 range is assembled on the page from Retell Voice Infra at $0.055/minute (transcription has no line of its own, so it appears folded in), the LLM at the model’s own per-minute rate from $0.0016/min for GPT 5 nano to $0.64/min for GPT 6 Astra on the fast tier, text-to-speech at $0.015/minute for platform voices or $0.040/minute for ElevenLabs, and telephony at $0.015/min.

The telephony line earns the position: it is priced, and it is removable, since the page states there is no charge for SIP trunking or custom telephony. There is no platform subscription, sign-up is self-serve with $10 in free credits, and every account includes 20 concurrent calls before $8.00/Concurrency/month. The page still never totals anything, and the 4.4x spread means the $0.07 floor only holds with the cheapest model and Retell’s own voices.

GOOD
  • Every layer of the minute is priced, including the carrier, and the carrier line drops to zero on your own SIP trunk.
BAD
  • Compliance is metered on top rather than included — PII removal adds $0.01/min, safety guardrails $0.005/min — and the Terms state all fees are non-refundable.

3. Deepgram Voice Agent API

Best for: One bundled rate for speech, model and voice · From: $0.075/min (Pay As You Go) / $0.068/min (Growth)

Why it sits at 3 $0.075/min bundles STT, model and voice; no telephony rate published

Deepgram’s Voice Agent API is the only rate here that puts speech-to-text, the model and the voice inside one published number: $0.075/min on Pay As You Go, $0.068/min on Growth. Bringing your own LLM and TTS cuts Deepgram’s side to $0.050/min, while frontier managed models raise it to $0.163/min, 2.17x the Standard rate. Sign-up is self-serve on a one-time $200 credit, with no minimums, no expiration and no credit card.

The carrier is the gap. The word telephony never appears on the pricing page, so the fourth layer of your minute comes from a third party you contract yourself. Two meters also matter: minutes bill on how long the websocket stays open, so silence and hold time are charged, and Pay As You Go is capped at 45 concurrent Voice Agent sessions, rising to 60 on Growth, with no published price for more below Enterprise.

GOOD
  • Three of the four cost layers sit inside one rate, and unused credits bought within the last 30 days are refundable — the only refund policy on this list.
BAD
  • No telephony price exists on the page at all, and billing runs on connection time, so dead air on a call costs the full per-minute rate.

4. Twilio ConversationRelay

Best for: Keeping the agent and the carrier on one bill · From: $0.07 / minute

Why it sits at 4 $0.07/min agent, $0.0140/min carrier, model billed by you

ConversationRelay is $0.07 / minute with no platform fee and no commitment, and because Twilio is also the carrier, the rest of the call sits on the same rate card: $0.0140 / min outbound local and toll-free, $0.0085 / min inbound local, and $1.15 / mo for a local number. The published 95th-percentile latency figure is under 0.725 seconds, the only percentile-qualified latency number in this comparison.

The model is the hole. Twilio expects you to bring and pay for your own LLM, so that cost never appears on a Twilio page, and the pricing page never resolves whether $0.07 includes speech-to-text and text-to-speech for ConversationRelay specifically. Optional components meter separately — Conversation Intelligence at $0.005 / 1K characters, Conversation Memory profiles at $0.005 per daily utilized profile after 100k free operations a month — and there are no free agent minutes, only a $15 trial credit published on a different page.

GOOD
  • Agent minutes and carrier minutes come from the same vendor on the same site, with pricing stamped current as of August 2026.
BAD
  • Twilio never states whether the $0.07 covers STT and TTS, and the model bill is entirely yours and entirely unpriced by Twilio.

5. Google Cloud Conversational Agents

Best for: Per-second billing inside Google Cloud · From: $0.001 / 1 second

Why it sits at 5 $0.001 per second, and no telephony rate for the current product

Google meters voice per second rather than per minute: $0.001 / 1 second for a Flows (Dialogflow CX) voice agent and $0.002 / 1 second for the generative Playbooks edition, with chat at $0.007 and $0.012 per request. There is no plan, no seat and no minimum, trial credits apply automatically on first use — $600 for Flows, $1,000 for Playbooks, expiring after 12 months — and building the agent is not metered at all.

The billing rules cost more than the rate implies. Both directions of a turn are charged and rounded up to the second, a caller barging in still bills the whole generated response, and a Playbook that calls a Flow prices the entire conversation at the Playbooks rate. The omissions are large: no telephony line item exists for Conversational Agents (the only published gateway rate is the older Dialogflow ES one at $0.05 / 1 minute tolled), and the page never says whether the per-second rate includes speech-to-text and text-to-speech, or which model Playbooks runs.

GOOD
  • Real trial credits of $600 or $1,000, with no subscription, seat fee or minimum anywhere in the product.
BAD
  • No telephony rate for the current product and no statement of what the per-second price bundles, so the all-in minute cannot be assembled from the page.

6. ElevenLabs (ElevenAgents)

Best for: A few hundred minutes a month, self-serve · From: Free $0 per month; Starter $6 per month; Creator $22 per month, first month $11

Why it sits at 6 15 minutes free monthly; extra minutes $0.080, burst $0.160

ElevenAgents sells minutes inside a subscription. Free is $0 per month with 15 minutes of calls and 4 concurrent calls; Starter is $6 per month with 75 minutes; Creator is displayed as $22 per month with the first month at $11, so $22 is the ongoing price and $11 buys month one only. Pro is $99 per month with 1,238 minutes, Scale $299 with 3,738 and Business $990 with 12,375.

Extra minutes are a flat $0.080 per minute on every tier, doubling to $0.160 under burst pricing when you exceed your plan’s concurrency — cheaper per minute than most self-serve options above. The rate bundles ElevenLabs’ own text-to-speech and speech-to-text, but not the model or the carrier: both appear only as At cost with no figure, LLM fees are deducted from your ElevenLabs credits at rates shown inside the product, and only one model surfaces publicly in the on-page calculator. The plan price is knowable; the minute is not.

GOOD
  • The only standing free plan here that runs real calls: 15 minutes a month with the workflow builder and knowledge base, not a countdown trial.
BAD
  • Telephony and the model are published as At cost with no number, and the Terms state all sales are final with prepaid credits non-refundable.

7. Bland AI

Best for: One flat rate with no model shopping · From: $0.14 /min (Start) — $0 platform fee. No card required.

Why it sits at 7 $0.14/min bundles the model; telephony priced nowhere

Bland publishes the simplest number in this comparison: $0.14 /min on Start with $0 platform fee and no card required, or $0.12 /min on Build, which adds a $299/month platform fee. Transfer time meters separately at $0.05 and $0.04 per transfer minute. The rate bundles the language model, speech-to-text and text-to-speech with no token charges and no provider pass-throughs, so there is no model roulette and no second invoice to reconcile.

Telephony is the exclusion, described only as pass-through cost with no figure — and Bland’s homepage FAQ contradicts its own pricing page by saying telephony is included, so the published record disagrees with itself on exactly the question this ranking asks. Sign-up grants 2 credits plus an inbound number valued at $15/mo, without defining a credit in dollars or minutes or saying whether it expires, and the Terms of Service contain no payments, billing, auto-renewal or refund section at all.

GOOD
  • A single rate covers the model, transcription and the voice on a tier with no platform fee and 99.9% uptime published.
BAD
  • The vendor gives two contradictory answers about whether telephony is in the rate, and never publishes a figure for it either way.

8. RingCentral AI Receptionist

Best for: A finished receptionist, not a platform · From: Starts at $49 /month

Why it sits at 8 100 minutes included, then $0.50/min in 30-second increments

This is a finished product rather than a platform to build on: $49/month standalone with 100 minutes included, a dedicated new phone number, unlimited AI Receptionists, transcripts and analytics, plus HIPAA and SOC2 — the only compliance listed inside a self-serve monthly price here rather than behind Enterprise. Bought with RingEX it reads $39 + $30 per month, a $69 combined entry the page never totals for you.

On this criterion it is the honest expensive option. The carrier, the number and transcription all sit inside the subscription, so the figure you read is close to the figure you pay — but minutes past the included 100 run at $0.50 per minute, and call time is rounded up in 30-second increments, so a call of one minute ten seconds bills as 1.5 minutes. Cheaper minute bundles are said to exist but are not priced, and the page adds that additional charges may apply.

GOOD
  • Carrier, number and transcription are inside one published monthly price, with HIPAA and SOC2 on the $49 plan.
BAD
  • Overage is $0.50 per minute with 30-second rounding, the minute bundles are unpriced, and the trial button appears only on the RingEX card despite the FAQ promising it on both.

9. Aircall AI Voice Agents

Best for: Teams already paying for Aircall seats · From: $0.37/min for up to 2,500 minutes, $0.29/min over 2,500 minutes

Why it sits at 9 $0.37/min self-serve, plus a $90/month three-licence floor

The AI Voice Agent is an add-on to a seat-based phone plan. Pay-as-you-go is $0.37/min up to 2,500 minutes and $0.29/min above, with prepaid bundles at 500 minutes for $130 and 2,500 for $550. The advertised from $0.14/minute requires 10,000+ minutes through a sales-negotiated custom bundle. Underneath sits a licence floor: Essentials is $30 /license, shown as 3 licenses for $90/month, billed annually, so the account costs $90 a month before a single AI minute.

Aircall’s own pages disagree with each other. The voice agent product page and the help centre list $0.49/min and $0.39/min against the pricing page’s $0.37 and $0.29, and the product page advertises no contract and no minimum usage against a pricing page with a three-licence minimum and a help centre stating you are committed for the period in your signed contract. Telephony is billed separately from the AI rate, refunds are affirmatively refused, and above 11 seats a cancellation filed inside 30 days of renewal renews the contract.

GOOD
  • Every plan carries 50 free AI voice minutes per account per month plus 100 at sign-up, with the limit documented in the help centre.
BAD
  • Two different per-minute rate cards are live on Aircall’s own site, and the monthly (non-annual) licence price is not published in readable form.

10. Synthflow

Best for: Enterprise buyers with a sales cycle · From: Enterprise contracts start at $30,000 annually.

Why it sits at 10 $30,000 annually is the only price on the page

Synthflow’s pricing page publishes exactly one number: enterprise contracts start at $30,000 annually. There is no second tier, no self-serve plan and no per-minute rate on it; the calls to action are CONTACT SALES and GET A DEMO. You can sign up and build an agent at no cost — call flows, voicemail detection, integrations — but launching moves you onto a pay-as-you-go model whose rate the page does not state.

It finishes last on both halves of the criterion. The all-in minute is not merely unbundled, it is unpublished: the only per-minute figure anywhere on Synthflow’s site sits in a blog post dated August 28, 2025 that neither the current pricing page nor the docs repeat, and the former self-serve pay-as-you-go docs page now redirects to enterprise pricing. Useful metering detail survives — per-second voice, a flat 5 seconds billed for unanswered calls, five AI chat messages to the minute — but none of it is a price.

GOOD
  • Metering rules are published in unusual detail, including per-second voice billing and no charge for genuinely failed calls.
BAD
  • No current per-minute rate exists on the site, and cancellation, renewal and package changes are all handled privately inside your agreement.

Which one fits your situation

You want to hear an agent answer a call this week and spend nothing

The free tier is enough. ElevenLabs Free is $0 per month with 15 minutes of calls and 4 concurrent calls as a standing plan, and Deepgram hands new accounts a one-time $200 credit with no minimums and no credit card. Neither needs a sales call.

You already buy numbers and minutes from Twilio

Twilio ConversationRelay at $0.07 / minute keeps the agent and the carrier ($0.0140 / min outbound, $0.0085 / min inbound, $1.15 / mo per local number) on one bill. Budget your model separately: Twilio expects you to bring and pay for your own LLM.

You run your own SIP trunk and want the lowest controllable minute

Retell. Voice Infra is $0.055/minute plus the model and voice you pick, and the $0.015/min telephony line disappears entirely, because the page states there is no charge for SIP trunking or custom telephony.

You need one number answered professionally and nobody to build it

RingCentral AI Receptionist standalone, $49/month with 100 minutes included, a dedicated number, HIPAA and SOC2 on the plan and a 14-day trial. Watch the $0.50 per minute overage and the 30-second rounding before you plan volume on it.

You want one rate and no model shopping

Bland Start at $0.14 /min with $0 platform fee bundles the model, speech-to-text and text-to-speech in a single number. Add your own carrier cost, because Bland prices telephony nowhere and its homepage contradicts its pricing page about whether it is included.

A vendor has quoted you a five-figure annual contract

Do not buy anything yet. Synthflow publishes only that enterprise contracts start at $30,000 annually, with no current per-minute rate on its pricing page or docs. Get the rate, the concurrency plan and the renewal handling in writing, then set it against the published self-serve rates above.

What this market gets people for

  • Treating a platform’s per-minute rate as the price of a minute Vapi’s $0.05/min is a hosting fee with models passed through at cost, and the page publishes no blended total. Retell lists every component and still never totals them. Twilio never says whether its $0.07 covers speech-to-text and text-to-speech. The figure in the pricing hero is almost never the figure on the invoice.
  • Budgeting minutes and forgetting concurrency Concurrency is metered separately nearly everywhere. Vapi’s $0 tier allows 4 concurrent calls, with extra lines at $10/line/month. Retell includes 20 and then charges $8.00/Concurrency/month. ElevenLabs doubles overage minutes from $0.080 to $0.160 under burst pricing when you exceed your plan’s concurrency. Deepgram hard-caps Pay As You Go at 45 sessions and publishes no price for more below Enterprise.
  • Planning around a from price that only sales can sell you Aircall advertises from $0.14/minute, reachable only at 10,000+ minutes through a sales-negotiated bundle; the self-serve rate is $0.37/min up to 2,500 minutes. Bland’s cheaper $0.12 /min arrives attached to a $299/month platform fee. Even Vapi’s cheapest subscription, Core at $29/month, carries a Contact Sales button rather than a checkout.
  • Assuming prepaid money comes back Retell’s Terms state all fees are non-refundable. ElevenLabs states all sales are final, with prepaid credits explicitly non-refundable. Aircall refuses pro-rated refunds for mid-cycle deletions. Vapi, Bland, Synthflow, Twilio, Google, AWS and RingCentral publish no refund window at all. Deepgram is the only vendor here with a stated policy: unused credits purchased within the last 30 days.

Questions buyers ask

What does a Vapi minute actually cost?

The $0.05/min headline is Vapi’s hosting fee only. The same page prices transcription (Deepgram) at $0.0095-$0.0099/min, the intelligence model (OpenAI) at $0.0077-$0.0452/min and voice (ElevenLabs) at $0.0146-$0.0238/min as pass-throughs at cost, then Twilio carrier at $0.008/min inbound or $0.014/min outbound. Vapi’s own WebSocket and WebRTC transports are free. No blended total is published anywhere on the page.

Which platforms include telephony in the per-minute price?

Essentially only the phone-system products. RingCentral’s $49/month standalone plan includes a dedicated number and 100 minutes inside the subscription. Retell ($0.015/min), Twilio ($0.0140/min outbound, $0.0085/min inbound) and Amazon Connect ($0.0048/min outbound, $0.0022/min inbound plus $0.0300 per day for a number) price the carrier as a separate published line. Deepgram, ElevenLabs and Bland publish no telephony rate at all.

Can I leave Vapi without talking to a salesperson?

On most of this list, yes. Retell, Deepgram, Twilio, Google, Bland, ElevenLabs, Amazon Connect, RingCentral (Buy now) and Aircall all show self-serve paths; only Synthflow requires sales for a paid deployment, and Aircall’s 10,000+ minute pricing does too. Worth noting on the way out: Vapi itself is self-serve only on the $0 prepaid tier — Core at $29/month, Pro and Premier all read Contact Sales.

Which free allowances actually run calls?

ElevenLabs Free gives 15 minutes of calls per month as a standing plan, and Aircall includes 50 AI voice minutes per account per month plus 100 at sign-up. The rest are credits, not minutes: Deepgram $200 one-time with no card required, Retell $10, Vapi $5, Bland two undefined credits, Google $600 for Flows or $1,000 for Playbooks expiring after 12 months. Amazon Connect publishes no free voice minutes at all.

Does anyone publish a renewal price that differs from the first price?

Only ElevenLabs, and only on one tier: Creator is $22 per month with the first month at $11. Every other price on this list is a single standing rate. That is not the same as nothing renewing — Retell’s Terms describe automatic renewal for the subscription period without publishing a figure, Vapi’s Core is marked prepaid and auto-renew, and Aircall automatically renews accounts above 11 seats if a cancellation lands inside 30 days of the renewal date.

Why isn’t every product in this category ranked here?

Three were set aside on purpose. Vapi is the benchmark this article measures against rather than an alternative to it. Rime sells text-to-speech by the character ($0.03 / 1K characters for Mist v3), not an agent you can put on a phone line. Microsoft Copilot Studio prices voice agents at 10, 35 or 75 Copilot Credits per minute and a credit at $0.01, but on two different pages, and never prints a dollar cost for a voice minute; real-time voice also requires a Dynamics 365 Contact Center licence and an Azure Communication Services number, neither priced on the Copilot Studio pages.

What none of them publish

These are the questions a buyer would want settled that no vendor on this page answers in public. If one of them decides your purchase, ask the vendor and hold them to the answer.

  • No vendor on this list publishes a blended all-in cost per minute. Even Retell, which prices every component separately, offers no worked example or calculator that totals them.
  • Telephony is missing entirely from Deepgram’s, ElevenLabs’ and Bland’s pricing pages — Bland calls it pass-through cost with no figure, ElevenLabs lists the provider only as At cost.
  • Whether Twilio’s $0.07 / minute and Google’s $0.001 / 1 second include speech-to-text and text-to-speech. Neither page says, and neither prints a separate line for the current product.
  • What a Bland credit buys: the page grants 2 credits at signup and never defines one in dollars or minutes, or says whether it expires. Vapi’s $5 credit and Retell’s $10 credit have no published expiry either, and Deepgram’s FAQ answer on what the $200 covers does not render on the page.
  • Enterprise and contract floors: Retell Enterprise, Bland Enterprise, Twilio’s contract rates, Deepgram Enterprise, ElevenLabs Enterprise, Aircall Custom (25-licence minimum) and RingCentral’s contact-centre AI Agents all publish no figure.
  • Annual pricing: ElevenLabs shows a Yearly billing toggle with no readable annual price, Aircall’s monthly (non-annual) per-licence price is not displayed on the default annual view, and RingCentral publishes no annual option for AI Receptionist at all.
  • Refund windows: only Deepgram publishes one, on unused credits purchased within the last 30 days. Retell and ElevenLabs affirmatively refuse refunds, Aircall refuses pro-rated credits, and the rest publish nothing.
  • Latency is published in incompatible forms and never as an SLA — Twilio’s 95th-percentile figure under 0.725 seconds, Retell’s ~600ms and Bland’s 400ms as homepage marketing, Synthflow’s under 100 ms for its own telephony network — while Google, Amazon Connect, RingCentral, Aircall and ElevenLabs publish no number at all.
  • Overage behaviour: what happens when Vapi’s prepaid credit runs out, when Bland hits its 100 calls/day Start cap, or when Deepgram exceeds 45 concurrent sessions. None of it is priced publicly.
  • Synthflow’s current per-minute rate, the price of RingCentral’s additional minute bundles, and the dollar cost of a Microsoft Copilot Studio voice minute.

Reading this from the other direction? AI Phone Answering Service for Small Business: 10 Platforms Ranked by Cost and Setup Effort covers the same market against a different rule.