Retell AI vs Vapi: which AI phone agent platform should you build on in 2026?
Last updated: November 9, 2026. This is independent research based on each vendor's published pricing and documentation, read on November 9, 2026. We did not test either product. See the full ranked list of AI phone agents.
Answer in brief
Pick Retell AI if you want the shortest path from signup to a live phone line. It has one published price range ($0.07–$0.31 per minute for voice agents), 20 concurrent calls free, billing to the nearest second, about 600 ms of advertised latency, and no telephony charge when you bring your own SIP trunk. Pick Vapi if your engineers want to choose every layer themselves. Vapi charges a $0.05/minute hosting fee, passes model costs through with no markup, lists 15 text-to-speech providers, and gives you free Vapi telephony/SIP, WebSocket and WebRTC transport. In short, Retell wins on included capacity and simplicity, and Vapi wins on choice of components.
Top pick for most phone teams: Retell AI. You get 20 free concurrent calls (Vapi's free tier has 4), a $10 starting credit, and a calculator that prices the LLM, voice and telephony per minute before you commit.
Best for builders who want every dial: Vapi.
Retell AI vs Vapi at a glance
| What matters on a call | Retell AI | Vapi |
|---|---|---|
| Published price | $0.07–$0.31/min for voice agents, depending on the LLM and voice you choose | $0.05/min hosting, plus transcriber, model, voice and transport at provider cost |
| Free start | $10 in credits, no contract | $5 in credits |
| Concurrent calls included | 20 free; $8 per extra call/month | 4 free (10 on Core, 30 on Pro); $10 per extra line/month |
| Advertised latency | About 600 ms (Retell home page) | About 800 ms end to end, typically (Vapi FAQ) |
| Voices | Retell platform voices with automatic TTS fallback, plus MiniMax, Cartesia, Fish Audio, Inworld and ElevenLabs | 15 listed TTS providers, including Vapi Voices, ElevenLabs, Cartesia, Deepgram, Hume, Rime AI, OpenAI and Azure |
| Telephony | Retell Twilio/Telnyx from $0.015/min (US); numbers $2/month; no charge for SIP trunking or custom telephony | Vapi telephony/SIP, WebSockets and Daily WebRTC are free; Twilio is $0.008/min inbound and $0.014/min outbound; Telnyx $0.0055/min; Vonage $0.00814/min |
| Billing | Tracked to the nearest second, no rounding per call | Pre-paid usage |
| HIPAA | Described as HIPAA-ready, with HIPAA/BAA options; SOC 2 certified | $2,000/month add-on that includes a BAA |
| Support | Community and email; 24/7 dedicated portal on Enterprise | Discord and email with no SLA; Core $29/mo (2-business-day SLA); Pro from $999/mo (99% uptime SLA); Premier 99.9% |
Latency: the half-second that decides whether callers stay
On a phone line, a slow reply sounds like a dropped call. Retell says independent benchmarks put it at about 600 ms. Vapi's FAQ says its end-to-end voice processing is typically around 800 ms. Both are vendor figures, and they may measure different things (time to first audio, full turn, different models). Treat them as a starting point, then place your own test calls with the exact model and voice you plan to run. On either platform, the LLM you pick moves latency more than anything else.
Voice quality: a curated menu or an open catalog
Retell offers its own platform voices from $0.015/min with automatic text-to-speech fallback, plus custom voices from MiniMax, Cartesia, Fish Audio, Inworld and ElevenLabs. ElevenLabs tiers run from $0.04/min (Flash) to $0.10/min (v3 and v4). Vapi's catalog is wider, with 15 providers including Hume, LMNT, Rime AI, WellSaid and xAI. If you want to A/B test many voices against your callers, Vapi gives you more options. If you want good defaults with a safety net, Retell's fallback is a real advantage.
Telephony: both speak SIP, and they price it differently
Retell supports elastic SIP trunking or SIP URI dialing, with step-by-step guides for Twilio, Telnyx and Vonage. Your own carrier costs nothing extra. Outbound extras are priced separately: batch calls cost $0.005 per dial and branded caller ID costs $0.10 per outbound call. Vapi's own telephony/SIP is free, and its status page tracks inbound and outbound service separately for SIP, Twilio, Telnyx and Vonage. That's useful if you run outbound campaigns at scale. Vapi includes 1 phone number on the free tier, 5 on Core and 10 on Pro.
Pricing model
Retell AI: pay as you go. With the calculator's default setup, the price is $0.150/min: $0.080 for the LLM, $0.055 for Retell voice infrastructure and $0.015 for TTS. Add-ons cost extra, for example a knowledge base (+$0.005/min), PII removal (+$0.01/min) and AI QA (first 100 minutes free, then $0.10/min). After a transfer, the AI fee stops and only telephony continues. Silence and hold time are billed. Failed calls are not.
Vapi: $0.05/min hosting plus provider costs. Vapi's own estimate for 1,000 minutes is about $82–$129/month on Vapi SIP. Success packages and HIPAA cost extra.
When to pick which
- Pick Retell AI if you need 10–20 simultaneous calls on day one, want one dashboard price per minute, or plan to bring an existing SIP trunk.
- Pick Vapi if you are building voice into your own product, want to swap transcribers, models and voices freely, or need WebRTC and WebSocket transport alongside phone calls.
- Look elsewhere if you just want a receptionist without engineering work. Start with Smith.ai vs Goodcall.
FAQ
Is Retell AI cheaper than Vapi?
It depends on your stack. Retell's default setup in its calculator is $0.150/min. Vapi estimates $82–$129 for 1,000 minutes, about $0.08–$0.13/min, before success packages or HIPAA. Price your exact model and voice on both calculators.
Which has lower latency, Retell AI or Vapi?
Retell advertises about 600 ms and Vapi says about 800 ms is typical. Those are vendor claims that may be measured differently, so test with your own calls.
Can I bring my own Twilio or SIP trunk?
Yes, on both. Retell charges nothing for SIP or custom telephony. Vapi supports Twilio, Telnyx, Vonage and custom SIP trunks.
How many simultaneous calls do I get for free?
Retell includes 20 concurrent calls. Vapi's free tier includes 4.
Keep comparing
Ranked list of the best AI phone agents · Retell AI review · Vapi review · Retell AI vs Bland AI · Vapi vs Bland AI · Alternatives to Retell AI · How to choose · Pricing comparison
Need hosted LLMs for a custom voice stack? See bestinferenceproviders.com. For chat and email support agents, see bestaisupportagent.com.
·
Sources
- Retell AI pricing
- Retell AI home page (latency claim)
- Retell AI custom telephony docs
- Retell AI custom voice providers
- Retell AI platform voices
- Vapi pricing
- Vapi FAQ (latency)
- Vapi voice providers
- Vapi SIP trunking
- Vapi status page
Named on this page as a provider? Join the network for a linked profile, Visit buttons and featured placement.