Vapi vs Retell
Two AI voice agent platforms, each priced by part, not one flat rate. One leads on inbound quality; the other on provider choice.
Choose Retell when inbound call quality and low latency matter most, like an AI receptionist taking live calls. Choose Vapi when you want to pick the model, voice and phone provider yourself, more than you want inbound calls tuned from day one.
The short answer
- Neither charges one flat per-minute rate. Both bill the voice agent apart from the model, text-to-speech, speech-to-text and phone lines. Each of those is its own line item.
- Retell is widely seen as stronger on inbound call quality and latency. You notice it most when a caller is waiting, like at a front desk.
- Vapi is widely seen as stronger on control for developers and agencies. A team can pick and swap its model, voice and phone providers more freely.
- Neither platform is a CRM. Whichever one runs the call, something else has to track what happened next, like whether the lead booked.
Vapi vs Retell at a glance
| Dimension | Vapi | Retell |
|---|---|---|
| Published starting rate | $0.05/min: platform hosting fee only | $0.055/min: floor for the voice layer |
| What that rate excludes | Model, voice, speech-to-text and phone lines, each billed at provider rates | Model, voice and phone lines, each billed as an add-on |
| Realistic all-in range | About $0.10–$0.30/min, by model and voice | About $0.07–$0.31/min, by model and voice |
| Where it's generally strongest | Control for developers and agencies: pick the model, voice and phone provider | Inbound call quality and latency |
| Provider flexibility | Broad. Swap the provider for each part | Possible, though it leans toward a set default stack |
| Best-fit use case | Teams that want to tune each layer themselves | Teams that want strong inbound calls without deep tuning |
Rates from each vendor's own pricing page, checked September 2026 (vapi.ai/pricing, retellai.com/pricing). Both leave model, voice and phone costs out of the starting rate. The all-in range comes from third-party cost breakdowns that add those parts on top. It is not either vendor's own quote.
What each platform does well
Vapi
- You choose the model, voice and phone provider behind the agent
- Popular with developers and agencies building custom voice flows
- The $0.05/min rate covers hosting only. Your real bill depends on the providers you add
- All that choice means more setup before the stack runs well
Retell
- Widely seen as strong on inbound call quality and latency
- No separate platform fee. Pricing is built around parts from the start
- The $0.055/min figure is a floor, not the full cost. The model and voice add on top
- Less focus on the deep provider swapping that Vapi is built around
How to choose
Start with what the calls are for. Say the agent answers inbound calls, and a real person waits on the line for a reply. Then every extra hundred milliseconds hurts. That's where Retell is most often strongest. Now say the calls are outbound and you're building a set flow. You want to pick the model that thinks and the voice that speaks. Then Vapi's choice of providers helps more.
Either way, price the whole stack before you compare headline rates. Each number on the pricing page covers only that platform's own layer. Add the model calls, the voice engine and the phone minutes. Then both land in the same rough range. Reports put it at about $0.07 to $0.31 a minute, by setup. That's well above the $0.05 or $0.055 each one leads with.
Whichever platform you pick, plan for a tool behind it to catch the result: how the call ended, the booking, the next task. A voice agent that answers well but saves the result nowhere is a call log, not a pipeline.
Pricing changes; verify before you commit
Questions
- Can you use both Vapi and Retell?
- Yes, and some teams do. They run one for outbound campaigns and the other for inbound. Or they trial both on the same job before they commit. Neither one locks you in. Both bill per part, not by a flat plan. So a small pilot on each is a fair way to compare real call quality before you choose.
- What does a voice agent actually cost per minute?
- More than either platform's headline number. Vapi's $0.05 per minute is its own hosting fee only. Retell's $0.055 per minute is a floor for its voice layer. Both then add the model, text-to-speech, speech-to-text and phone lines on top, each billed by use. So the price on the page is a start, not the bill. Checked on each vendor's pricing page in September 2026; see the table note below.
- Which platform has lower latency?
- Retell is widely seen as stronger on latency and inbound call quality. That matters most when a caller is waiting on the line, like at a front desk. Vapi's strength lies elsewhere. It lets you choose which providers plug into the stack.
- Do I need a CRM behind either platform?
- Yes, in practice. Vapi and Retell both run the voice agent: the part that listens, decides and speaks. Neither one tracks what happened after the call. Was the lead booked? How did the call end? What's next? You need a separate tool for that, whichever voice platform you pick.