TL;DR / Key Takeaways
- At 1,000 minutes a month, the main platforms cost roughly $75 to $140 before phone charges.
- ElevenLabs Agents includes minutes in every plan and charges $0.08 per extra minute, plus the LLM.
- Retell has the clearest per-minute pricing and 20 free concurrent calls.
- Vapi passes providers through at cost but allows only 4 concurrent calls on pay-as-you-go.
- Bland bundles the LLM into one rate: simplest bill, highest per-minute price.
Short answer: for most teams building a phone or web voice agent, the choice is between four platforms. ElevenLabs Agents has the best-known voices and includes call minutes in every plan (extra minutes are $0.08). Retell has the clearest per-minute pricing and 20 free concurrent calls. Vapi gives developers the most control and passes provider costs through at cost. Bland bundles the AI into one per-minute rate, which makes the bill simple but higher. At 1,000 minutes a month, all four land between roughly $75 and $140 before phone charges. Updated October 29, 2026.
How we checked: every price, limit and plan below comes from the vendor's own pricing page or documentation, read October 29, 2026. Latency and compliance figures are the vendors' own claims; we did not test them. The monthly cost table is our own arithmetic on those published rates, and its assumptions are listed under it.
The platforms at a glance
| Platform | How you pay | Free to start | Concurrent calls | Phone line |
|---|---|---|---|---|
| ElevenLabs Agents | Plan with included minutes; $0.08/min after that; LLM billed separately at the provider's rate | Free plan: 15 minutes | 4 (Free) to 40 (Business) | Bring your own: Twilio or SIP trunk, no ElevenLabs telephony fee |
| Retell AI | $0.055/min platform + voice + LLM + phone, each per minute | $10 credit | 20 free; $8/month per extra | Twilio or Telnyx at $0.015/min, or your own SIP for free |
| Vapi | $0.05/min hosting + providers at cost | $5 credit | 4 on pay-as-you-go; 10 on Core ($29/month) | Twilio, Telnyx or Vapi SIP, at cost |
| Bland AI | One rate covering LLM, speech-to-text and voice: $0.14/min (Start) or $0.12/min (Build, $299/month) | Start plan, $0/month | 10 (Start); 50 (Build) | Your carrier or Bland's at pass-through cost |
| Deepgram Voice Agent API | One rate: $0.075/min Standard; less if you bring your own LLM | $200 credit | Not stated on the pricing page | Not included |
What 1,000 and 10,000 minutes a month cost
| Platform | 1,000 minutes | 10,000 minutes | Not included |
|---|---|---|---|
| ElevenLabs Agents | About $80 (Starter or Creator plus extra minutes) | About $800 (Pro or Scale plus extra minutes) | LLM, phone line |
| Retell AI | $78 to $110 | $780 to $1,100 | Phone line ($15 or $150 via Retell) |
| Vapi | $82 to $129 (Vapi's own estimate) | About $820 to $1,290 | Plan fee if you need more than 4 concurrent calls |
| Bland AI | $140 (Start) | About $1,500 (Build: $299 plus $0.12/min) | Phone line |
| Deepgram Voice Agent API | $75 | $750 | Phone line |
- ElevenLabs: extra minutes at $0.08 on top of the plan's included minutes. The LLM is passed through at the provider's published price; as a reference point, Retell lists a small model (GPT-5 mini) at $0.008/min and a larger one (GPT-5) at $0.04/min.
- Retell: $0.055 platform plus $0.015 standard voice plus an LLM from $0.008 (GPT-5 mini) to $0.04 (GPT-5) per minute.
- Vapi: the 1,000-minute figure is Vapi's own estimate including providers; the 10,000-minute figure is that estimate times ten.
- Bland: at 10,000 minutes the Start plan's 100-calls-a-day limit is likely to bind, so we priced the Build plan.
- Silence and hold time count as billed minutes on Retell, and probably elsewhere: shorter, tighter calls are the biggest saving on every platform.
ElevenLabs Agents
ElevenLabs' agent platform (the product formerly called Conversational AI, now ElevenAgents) is the pick when the voice itself matters: a brand voice, a cloned voice, or many languages. Every plan includes minutes: 75 on Starter ($6), 275 on Creator ($22), 1,238 on Pro ($99), 3,738 on Scale ($299) and 12,375 on Business ($990). Concurrency rises with the plan, from 6 calls on Starter to 40 on Business, and you can burst to three times your limit at $0.16 per minute.
- LLMs: OpenAI, Anthropic, Google Gemini, ElevenLabs-hosted open models, or your own endpoint, passed through at the provider's price with no markup (some Gemini and Claude models are billed at Vertex rates, 10% higher).
- Building: knowledge base, tools, a visual workflow builder, automated tests and simulations, and analytics.
- Channels: native Twilio, SIP trunking, batch outbound calls, and connectors for Genesys, Amazon Connect, Zendesk and Salesforce.
- Compliance (as stated): SOC 2 Type II, ISO 27001, PCI DSS Level 1, HIPAA and GDPR. A HIPAA BAA is Enterprise-only.
Watch out for one thing: ElevenLabs' marketing page says 90+ languages while its agent docs list 31, so test your language first. If you plan to use your own voice, see our ElevenLabs voice cloning guide.
Retell AI
Retell is the easiest to budget for: every component has its own per-minute price, billing is per second, and failed calls are free. You get 20 concurrent calls at no charge, which is more than any other platform here gives on its entry tier. Retell states "~600ms latency" and HIPAA, SOC 2 Type II, GDPR and ISO 27001 compliance, with the BAA on Enterprise. Add-ons cost extra: knowledge base +$0.005/min, PII removal +$0.01/min.
Vapi
Vapi is built for developers who want to pick every provider (speech-to-text, LLM, voice, phone) and pay those providers at cost on top of a $0.05/min hosting fee. It states "sub-500ms latency" and SOC 2, HIPAA and PCI compliance. Two costs catch teams out: pay-as-you-go allows only 4 concurrent calls (Core at $29/month gives 10), and HIPAA is a $2,000/month add-on.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
Bland AI
Bland charges one rate that already includes the LLM, speech-to-text and voice: $0.14/min on the free Start plan or $0.12/min on Build at $299/month. That is the simplest bill here and the highest per-minute price. Start caps you at 100 calls a day. Bland lists a 99.9% uptime SLA and SOC 2, ISO 27001 and GDPR, with a HIPAA BAA on Enterprise.
Also worth a look
- Deepgram Voice Agent API: $0.075/min all-in on Standard, $0.065/min if you bring your own LLM, and a $200 credit that does not expire. Developer-first.
- Synthflow: sells annual Enterprise contracts only and does not publish prices.
- OpenAI Realtime API: priced per audio token ($32 per million input and $64 per million output tokens for gpt-realtime), not per minute, so compare it by running a real call.
Which one to choose
- You care most about voice quality, a cloned voice or many languages: ElevenLabs Agents.
- You want a predictable per-minute bill and plenty of concurrency from day one: Retell.
- You are a developer who wants to swap providers and tune every component: Vapi or Deepgram.
- You want one number per minute and do not want to manage an LLM bill: Bland.
Frequently asked questions
How much does an AI voice agent cost per minute?
On published 2026 rates, roughly $0.075 to $0.14 per minute before phone charges: Deepgram's Voice Agent API is $0.075, ElevenLabs Agents is $0.08 per extra minute plus the LLM, Retell is about $0.078 to $0.11 depending on the LLM, and Bland is $0.12 to $0.14.
Does ElevenLabs Agents include phone numbers?
ElevenLabs charges no telephony fee: you connect your own Twilio account or SIP trunk, and that provider bills you directly.
Which AI voice agent platform is HIPAA compliant?
ElevenLabs, Retell, Vapi and Bland all state HIPAA compliance. ElevenLabs, Retell and Bland offer the BAA on Enterprise plans; Vapi sells HIPAA as a $2,000-a-month add-on.
Can I use ElevenLabs voices in Retell or Vapi?
Yes. Retell lists ElevenLabs voices at $0.04 to $0.10 per minute and Vapi passes ElevenLabs voice costs through at $0.0146 to $0.0238 per minute, both on top of their platform fees.