Blog › Head-to-head
Synthflow vs Retell vs Vapi
All three are developer platforms for building voice AI, not finished receptionist products. You get APIs, SDKs and a dashboard; you build the agent, the prompts, the booking integration, the data model and the operations around it.
Vapi charges $0.05/min for hosting on its Build plan, with STT, LLM and TTS passed through at cost — or free if you bring your own API keys. 10 concurrent lines included, $10/line/month beyond. Note the Build plan retains call history for 14 days.
Retell is pay-as-you-go at $0.07–$0.31/min all-in, with the range driven by which LLM and voice you pick. 20 concurrent calls and $10 of free credits at signup.
Synthflow has moved upmarket: its pricing page now states enterprise contracts start at $30,000 annually. It is no longer competing for small-business budgets at all.
If you are a developer or agency building a product, Vapi and Retell are excellent and this is the right category for you. If you run a plumbing company or a dental practice and want your phone answered, you are looking at the wrong shelf — a finished AI receptionist costs $49–$149/month with none of the build work.
These are toolkits, not products
The most common mistake in this comparison is treating Vapi, Retell and Synthflow as alternatives to Rosie, Goodcall or Smith.ai. They are not in the same category, and the per-minute rates look deceptively cheap because of it.
A finished AI receptionist is something you point your phone number at. A voice AI platform is something an engineer builds one with. The per-minute rate buys you speech-to-text, an LLM turn loop, text-to-speech, telephony plumbing and a latency-optimised orchestration layer. It does not buy you a working receptionist for your business.
That distinction is the whole page. Everything below is either pricing detail or an accounting of the gap between those two things.
What each one charges
Vapi
| Item | Build plan |
|---|---|
| Vapi hosting | $0.05 / min |
| SMS / chat | $0.005 / msg |
| Concurrency | 10 included, then $10 / line / mo |
| STT, LLM, TTS | Passed through at cost, or free with your own keys |
| Call history retention | 14 days (30 days chat) |
| Support | Community Discord and email |
| SSO, RBAC, SOC 2 | Not on Build |
| HIPAA add-on | $2,000 / mo |
| Zero data retention add-on | $1,000 / mo |
Source: vapi.ai/pricing, read 29 July 2026. Vapi also offers a Scale plan on an annual contract with a fixed platform fee, committed volume and volume-based per-minute rates, adding SOC 2, HIPAA, PCI, SSO, RBAC, data residency, a dedicated account team and a custom support SLA.
Retell
| Component | Rate |
|---|---|
| Voice agents, all-in | $0.07 – $0.31 / min |
| Retell voice infrastructure | $0.055 / min |
| LLM — GPT 4.1 | $0.045 / min |
| LLM — Claude 4.6 Sonnet | $0.08 / min |
| LLM — GPT 5.5 | $0.16 / min |
| TTS — Retell platform voices | $0.015 / min |
| TTS — ElevenLabs voices | $0.040 / min |
| Telephony (US Twilio) | $0.015 / min |
| Knowledge base add-on | +$0.005 / min |
| PII removal add-on | +$0.01 / min |
| Phone number | $2.00 / mo |
| Concurrency | 20 included, then $8 / concurrency / mo |
| Free credits | $10 at signup |
Source: retellai.com/pricing, read 29 July 2026. Chat agents are listed at $0.002+/msg, advanced denoising at +$0.005/min, branded calls at +$0.10 per outbound call, and verified phone numbers at $10/month. Enterprise pricing is negotiated per client with dedicated support and infrastructure. The $0.07–$0.31 range is Retell's own published span; your actual rate depends on the LLM and voice you select from the component list above.
Synthflow
| Item | Enterprise |
|---|---|
| Minimum contract | $30,000 / year |
| Self-serve plans | Not published |
| Pricing basis | Scoped per customer |
Source: synthflow.ai/pricing, read 29 July 2026. The page states "Enterprise contracts start at $30,000 annually. Final pricing is scoped around call volume, concurrency, telephony setup, integrations, security needs, and launch support," and includes enterprise SLA and support terms, native telephony or SIP trunking, custom concurrency planning and routing, CRM/calendar/contact-centre integrations, MSA and DPA support, enterprise security review, and implementation and launch support. No monthly self-serve tiers, included minutes, per-minute overage rates, assistant counts or white-label fees were published at the time of reading.
Real cost at real volume
Per-minute rates are hard to feel. Here is the same call volume priced across all four options, using a three-minute average call.
| Monthly volume | Vapi platform only |
Retell at $0.12/min |
Synthflow | A finished product |
|---|---|---|---|---|
| 500 min · ~165 calls | $25 + models | $60 | $2,500 | $79 |
| 1,000 min · ~330 calls | $50 + models | $120 | $2,500 | $79 |
| 3,000 min · ~1,000 calls | $150 + models | $360 | $2,500 | $79 |
| Engineering to build it | Yours | Yours | Included | Included, free |
| Ongoing maintenance | Yours | Yours | Contracted | Included, free |
Our own arithmetic from the published rates above, assuming a three-minute average call. Vapi's column is the $0.05/min hosting fee only — STT, LLM and TTS are additional, either passed through at cost or billed to your own provider accounts if you bring your own keys, so the true figure is meaningfully higher. Retell's column uses $0.12/min, a mid-range estimate within its published $0.07–$0.31 span; your rate depends on LLM and voice selection. Synthflow's column is the $30,000 annual minimum divided by 12; actual pricing is scoped per customer and volume-independent at this floor. "A finished product" is WarmLane at launch pricing for one voice agent with unlimited inbound calls, plus your own Twilio and OpenAI usage billed at cost — see warmlane.io/pricing. Concurrency charges are excluded from all columns.
Read that table carefully, because the naive conclusion is wrong in both directions.
Vapi and Retell are genuinely cheap infrastructure. At 1,000 minutes a month, Retell at roughly $120 and Vapi at $50 plus model costs are reasonable prices for what they provide. Nobody is being gouged.
But the last two rows are where the money actually is. "Yours" is not free. It is engineering time — the most expensive input in the whole comparison, and the only one that does not appear on any pricing page.
The Synthflow change
Synthflow used to be discussed alongside Vapi and Retell as a more no-code-friendly option for smaller teams. That is no longer accurate, and it is worth stating clearly because a lot of comparison content has not caught up.
As of the reading date, Synthflow's published floor is $30,000 a year — around $2,500 a month — with no self-serve tiers listed. What you get for it is genuinely enterprise: SLAs, SIP trunking, custom concurrency and routing design, contact-centre integrations, MSA and DPA support, security review, and hands-on implementation.
That is a coherent product for a contact centre or a large multi-site operator. It is simply not a comparison point for a small business, an agency testing an idea, or a developer building a side project. If you arrived here from a two-year-old listicle that grouped all three together, that grouping has expired.
What you still have to build
This is the part the per-minute pricing hides. Starting from Vapi or Retell, a production AI receptionist still needs all of the following, and none of it is included:
- The knowledge layer. Your services, hours, pricing, policies, service area and FAQs, structured so the agent answers from ground truth instead of improvising. Getting this wrong is how AI agents invent prices.
- Prompt and persona engineering. Greeting, tone, qualification flow, objection handling, what to do when the caller goes off-script. This is iterative work measured in weeks of listening to real calls, not an afternoon.
- Booking integration. Function calling into Google Calendar or Calendly, reading genuine availability, respecting working hours, duration and daily caps, and not double-booking. This is the single largest chunk of build work.
- Data and storage. Transcripts, lead records, search, export. Note Vapi's Build plan retains call history for 14 days, so if you want a durable archive you are building and hosting that yourself.
- Telephony operations. Number provisioning, routing rules, business-hours behaviour, warm transfer, failover when a call drops, callback on missed calls.
- Monitoring and quality. Latency tracking, failure alerting, call review, and a way to tell whether last week's prompt change made things better or worse.
- A dashboard, if anyone other than you needs to see any of it.
For an engineering team building a product to sell, all of that is the point — it is your product, and Vapi or Retell is exactly the right foundation. We would genuinely recommend them for that.
For a plumbing company, a dental practice or a law firm, all of it is overhead standing between you and a phone that gets answered. And it is overhead that never really ends, because the agent needs maintaining as your business changes.
Before you scope a build, hear what a finished one sounds like.
This is our live production agent — AI Brain, booking, transfers, transcripts, all of it built and running. If it does what you were planning to spend six weeks building, that's worth knowing before you start.
+1 (218) 683-6315$79/month, unlimited calls, built and tuned for you at no extra cost.
Which to choose
You want the lowest platform fee and maximum control, and you already have API keys for your preferred STT, LLM and TTS providers — bringing your own keys makes the model layer free of Vapi markup. Best for engineering teams comfortable owning the whole stack. Watch the 14-day call history retention on Build, and the $2,000/month HIPAA and $1,000/month zero-data-retention add-ons if you are in a regulated space.
You want a more bundled all-in per-minute rate with telephony included, published component-level pricing so you can model costs precisely, and higher default concurrency (20 versus Vapi's 10). The $10 in free credits makes it the cheapest way to prototype. Good for agencies and product teams who want predictability without assembling every layer.
You are an enterprise or contact centre with a budget above $30,000 a year and requirements around SLAs, SIP trunking, security review and hands-on implementation. Not a fit for small business or self-serve.
You are the business that needs its phone answered, not the team building the tool. If your goal is bookings on your calendar rather than a system you own, the build work on these platforms is cost without benefit. WarmLane is $79/month with unlimited inbound calls, and our team builds and tunes the agent for you — typically live in about 30 minutes, same day.
Frequently asked questions
Is Vapi or Retell cheaper? +
How much does Synthflow cost? +
Can I build my own AI receptionist on these platforms? +
How long does Vapi keep call history? +
Do these platforms include phone numbers and telephony? +
What's the concurrency limit on each? +
Should a small business use Vapi, Retell or a finished AI receptionist? +
Or skip the build entirely.
Unlimited calls, booking, transfers, transcripts and AI analysis — built and tuned for you, live the same day.
+1 (218) 683-631514-day free trial, no card. Free Done-For-You setup and support included.