Best AI Voice Agent Platforms (September 2026), Decided on Published Pricing
The headline per-minute price is not the price of a call. Every platform here bills by the connected minute, but each one draws the line in a different place, and the language model, the transcription, the voice and the phone line may all sit outside the quoted figure. ElevenAgents is the easiest place to start: it is the only one with a standing free plan, at 15 minutes a month, and its Starter tier is $6 a month with 75 minutes included. Bland AI publishes the most complete minute: $0.14 on Start, stated to include the model, transcription and voices with no token charges. Vapi is the cheapest shell at $0.05 a minute for hosting alone, which suits you only if you already hold your own model and voice keys.
See ElevenAgents plans →The four, side by side
| Feature | ElevenAgents 🏆 | Vapi | Retell AI | Bland AI |
|---|---|---|---|---|
| Free tier | Yes: 15 min/mo, 4 concurrent ✅ | Not stated | $10 in credits | 2 credits + inbound number, no card |
| Cheapest way in | Starter $6/mo, 75 min included ✅ | Build: no platform fee, $10 per line/mo | Pay as you go, starts at $0 | Start: $0 platform fee |
| Quoted rate | $0.080/min overage | $0.05/min hosting ✅ | $0.07-$0.31/min headline | $0.14/min talk time |
| What the rate covers | TTS and STT yes; LLM and telephony billed at cost | Hosting only; STT, LLM and TTS at cost, or $0 with your own keys | Priced in parts: $0.055 infra + $0.015 TTS + $0.003-$0.16 LLM + $0.015 US telephony | LLM, STT and TTS all stated included, no token charges ✅ |
| Concurrent calls | 6 on Starter | 10 on Build | 20 included ✅ | 10 on Start |
Prices are the vendors' USD figures as printed. ElevenAgents and Bland AI state no annual pricing; Vapi's Scale tier and Retell AI's Enterprise tier are both custom. Sources and dates are under each card below.
How we decide
Voice agent platforms are decided on four published facts: how much of a call the quoted rate actually covers, what the cheapest way in costs, how many concurrent calls that tier allows, and whether you can try it without a card.
- The ranking rule for this page: Rank by how complete the quoted minute is, then by the cost of the cheapest tier that includes minutes, then by concurrency on that tier, then by whether a free tier exists; custom and enterprise-only pricing cannot be ranked and is noted instead.
- Where the facts come from: each product's own pricing and documentation pages, linked under every card with the date we checked them. When a vendor changes a plan, the fact changes here on the next check, and the date moves.
- What we do not do: star ratings, review counts, or claims to have tested the software in a lab. We compare what the companies publish, and we say when two products are close.
- How we are paid: some links pay us a commission if you start a paid plan. That is disclosed in the footer, it never changes the price you pay, and products with no partner program are ranked on the same rule as products with one.
The picks
ElevenAgents
The only standing free plan here, and $6 a month buys 75 minutes
- What the per-minute price includes The platform, its own text to speech and speech to text, RAG and the widget. The page states the language model and telephony are "billed separately at cost", so the quoted minute is not the whole cost of a call
- Free plan Yes: $0, 15 minutes per month, 4 concurrent calls
- Starter $6 per month: 75 minutes, 6 concurrent calls
- Creator $22 per month, shown with "First month 50% off" at $11: 275 minutes, 10 concurrent calls
- Pro $99 per month: 1,238 minutes, 20 concurrent calls
- Scale and Business Scale $299 per month: 3,738 minutes, 30 concurrent. Business $990 per month: 12,375 minutes, 40 concurrent. Enterprise is custom
- Overage $0.080 per minute, the same rate on every plan including Free
- Going over concurrency Burst pricing at $0.160 per minute, double the standard overage rate
- Text messages $0.003 per message
- Annual pricing Not stated on the agents pricing page
- Included on all paid plans Workflow Builder, Knowledge Base, multilingual support, widget, text to speech, speech to text, RAG, telephony
Facts from the ElevenAgents pricing page, checked 2026-09-08.
ElevenAgents is the only platform of the four with a free plan rather than a credit grant: $0 for 15 minutes a month and 4 concurrent calls. Starter is $6 a month for 75 minutes and 6 concurrent, Creator is $22 a month for 275 minutes and 10 concurrent with a first-month half-price promotion shown at $11, and Pro is $99 for 1,238 minutes. Overage is $0.080 a minute on every tier including Free, and exceeding your concurrency limit triggers burst pricing at $0.160, exactly double. Every paid plan lists the Workflow Builder, Knowledge Base, multilingual support, the widget, RAG and telephony. The important caveat is on the same page: the language model and telephony are billed separately at cost, so $0.080 is the platform's share of a minute and not the whole of it. No annual pricing is stated.
Bland AI
$0.14 stated to include the model, transcription and voices
- What the per-minute price includes The page states the rate covers "LLM, No token charges, STT Real-time transcription, TTS Premium voices". This is the only tool here whose quoted minute includes the language model
- Start plan $0 platform fee, $0.14 per minute of talk time and $0.05 per minute of transfer time, 10 concurrent calls, 100 calls a day
- Build plan "$299/month platform fee", $0.12 per minute talk time and $0.04 per minute transfer, 50 concurrent calls, 2,000 calls a day
- Free trial Start provides "2 credits + an inbound number ($15/mo value)" with "No card required"
- Enterprise Custom price, concurrency "sized to your volume", unlimited daily cap, on-prem or VPC available, forward deployed engineer available
- Annual pricing Not stated on the pricing page
Facts from Bland AI's pricing page, checked 2026-09-08.
Bland AI is the only one here whose quoted rate covers the language model. Its page states the price includes "LLM, No token charges, STT Real-time transcription, TTS Premium voices", which makes $0.14 a minute on the Start plan directly comparable to a number that other vendors reach only after adding two or three line items. Start has no platform fee, 10 concurrent calls, a cap of 100 calls a day, and a trial of 2 credits plus an inbound number the page values at $15 a month with no card required. Build adds a $299 a month platform fee and brings the rate down to $0.12, with 50 concurrent calls and 2,000 a day. Transfer time is billed lower than talk time, at $0.05 on Start and $0.04 on Build. It ranks second rather than first only because there is no free tier and no annual pricing stated.
Retell AI
Publishes each component of the minute separately, and includes 20 concurrent calls
- What the per-minute price includes The page breaks the minute into parts rather than bundling it: Retell voice infrastructure $0.055 per minute, text to speech $0.015 per minute, the language model $0.003 to $0.16 per minute depending on the model, and telephony $0.015 per minute on US Twilio with no charge for custom SIP
- Headline rate "$0.07-$0.31 / min for AI Voice Agents"; "$0.002+/ msg for AI Chat Agents"
- Free credits "$10 in free credits" on pay-as-you-go signup. "First 100 minutes free" for the AI Quality Assurance feature
- Pay as you go Starts at $0 with usage-based billing
- Concurrency Pay as you go: "20 included, more available on demand", additional capacity "$8.00/Concurrency/month". Enterprise: "No Cap on Concurrent Calls"
- Add-ons Knowledge Base $0.005/min, Advanced Denoising $0.005/min, Safety Guardrails $0.005/min, PII Removal $0.01/min
- Enterprise Custom pricing for both voice and chat agents
- Volume discounts Not stated on the pricing page
Facts from Retell AI's pricing page, checked 2026-09-08.
Retell AI does something the others do not: it prints the parts. Its infrastructure is $0.055 a minute, text to speech is $0.015, the language model is $0.003 to $0.16 depending on which one you choose, and US telephony over Twilio is $0.015 with no charge for a custom SIP connection. That is where the headline range of $0.07 to $0.31 a minute comes from, and it is honest about the spread rather than quoting the floor. Pay as you go starts at $0 with 20 concurrent calls included, the most of any entry tier here, with extra capacity at $8 per concurrent call a month. Signup comes with $10 in free credits. Optional extras are priced per minute too: knowledge base, denoising and guardrails at $0.005 each, PII removal at $0.01. No volume discounts are stated.
Vapi
$0.05 a minute for hosting, and nothing else included
- What the per-minute price includes Vapi hosting only. The page states "Model Provider Cost (STT, LLM, TTS): At cost ($0 if you bring your own API key)", so the speech, transcription and language model are extra unless you supply your own keys
- Base rate "Calls: $0.05 / min" for Vapi hosting. "SMS/Chat: $0.005 / msg"
- Build plan Usage-based, "60+ minutes included", support is "Community Discord, Email"
- Concurrency "10 call concurrency ($10 per line/month)" on Build
- Scale plan Annual contract with a fixed platform fee and committed volume, volume-based per-minute pricing, SOC 2, HIPAA, PCI, SSO and RBAC, dedicated account team. Price not stated
- Compliance add-ons HIPAA at $2,000 per month; Zero Data Retention at $1,000 per month
- Data retention Build: call history 14 days, chat history 30 days. Scale: custom
- Free trial Not stated on the pricing page
- Minimum spend Not stated on the pricing page
Facts from Vapi's pricing page, checked 2026-09-08.
Vapi is the thinnest layer and prices accordingly. Hosting is $0.05 a minute, chat is $0.005 a message, and the page states the model provider cost for speech to text, the language model and text to speech is charged at cost, or $0 if you bring your own API key. That makes Vapi the cheapest option for someone who already pays for those services directly and the hardest to price for someone who does not, because the pricing page carries none of those downstream numbers. Build has no platform fee but charges $10 per line a month for its 10 call concurrency, and keeps call history for 14 days. Scale is an annual contract with a fixed platform fee, committed volume and enterprise compliance, at a price not stated. Compliance is expensive if you need it: HIPAA is $2,000 a month and zero data retention is $1,000.
Which one should you actually get?
- You want to try one this afternoon without a card: ElevenAgents free plan, 15 minutes a month, or Bland AI Start, which includes an inbound number.
- You want one number you can forecast against: Bland AI at $0.14 a minute, because the model and transcription are inside it.
- You already pay for your own model and voice APIs: Vapi at $0.05 a minute, which charges you for hosting and nothing else.
- You expect bursts of simultaneous calls: Retell AI includes 20 concurrent on its entry tier, against 6 on ElevenAgents Starter.
- You want to see every cost before committing: Retell AI, the only one that publishes the minute as its components.
One thing every page here leaves out
None of the four state the total cost of a finished call in one number, because none of them control all four parts of it. A minute of conversation needs a language model, a transcription of what the caller said, a voice to answer with, and a phone line to carry it. Read any quoted rate as a question rather than an answer: which of those four does this include, and where do I pay for the rest? That is the first fact recorded on each card above, and it is the fact that changes the ranking.
Frequently Asked Questions
It depends on how much of the call the quoted rate covers, and on these four pages it covers very different things. Bland AI quotes $0.14 a minute on its Start plan and states that rate includes the language model, real-time transcription and premium voices, with no token charges. ElevenAgents charges $0.080 a minute in overage and includes its own text to speech and speech to text, but states the language model and telephony are billed separately at cost. Vapi charges $0.05 a minute for its hosting alone and bills speech, transcription and the model at cost on top, or nothing if you bring your own API keys. Retell AI publishes the parts separately: $0.055 for its infrastructure, $0.015 for text to speech, $0.003 to $0.16 for the model and $0.015 for US telephony. So the cheapest headline number is not the cheapest call.
ElevenAgents is the only one of the four with a standing free plan: $0 for 15 minutes a month and 4 concurrent calls. The others offer credits rather than a plan. Retell AI gives $10 in free credits on pay-as-you-go signup, and Bland AI Start gives 2 credits plus an inbound number it values at $15 a month, with no card required. Vapi does not state a free trial on its pricing page.
ElevenAgents Starter at $6 a month for 75 included minutes is the lowest committed price with minutes attached. Bland AI Start and Retell AI pay-as-you-go both have a $0 platform fee and charge only for what you use, which is cheaper if your volume is very low and more expensive per minute once it is not. Vapi Build has no platform fee either but adds $10 per line per month for its 10 call concurrency.
Concurrency is how many calls the agent can handle at the same time. Callers beyond that limit are queued or rejected, so it caps your busiest minute rather than your monthly total. On the cheapest paid tiers here: Retell AI includes 20, Vapi Build includes 10 at $10 per line per month, Bland AI Start includes 10, and ElevenAgents Starter includes 6. ElevenAgents charges burst pricing at $0.160 a minute, double its normal overage, when you exceed the limit.
None of the four state that telephony is included in the base rate. ElevenAgents states telephony is billed separately at cost. Retell AI prices US Twilio telephony at $0.015 a minute and charges nothing extra for a custom SIP connection. Vapi states transport costs are charged by the provider rather than by Vapi. Bland AI includes an inbound number with its Start plan but does not state that outbound telephony is included in the per-minute rate.
Every figure on this page was read from the vendor's own pricing page on 2026-09-08 and each card links to its source. This category changes faster than most, and two of the four publish no annual pricing at all, so the linked page is the authority.