No free plan. The cheapest way in is Build (usage-based) at $0.05/min Vapi hosting. Consider free alternatives in the voice ai category if budget is tight.
Vapi charges $0.05/minute platform fee plus underlying provider costs. A typical setup with Deepgram STT + GPT-4 + ElevenLabs TTS + Twilio telephony costs $0.15-$0.25/minute total. Premium voices and reasoning-heavy models can push costs to $0.33+/minute. The $10 free trial lets you test real costs before committing.
Vapi is more developer-oriented with flexible component selection (choose your STT/LLM/TTS providers), while Retell AI offers simpler setup with flat $0.07/minute pricing. Vapi gives more control and customization; Retell AI is easier to start with and has more predictable costs. Choose Vapi if you need deep customization, Retell for faster deployment.
No, Vapi is cloud-only. The real-time voice infrastructure requires specialized edge deployment for latency optimization. For self-hosted voice AI, you'd need to assemble components individually (Twilio + Deepgram + ElevenLabs + custom orchestration). Enterprise plans offer HIPAA compliance and dedicated infrastructure within Vapi's cloud.
Vapi provides SDKs for JavaScript/TypeScript (web and Node.js), Python, and REST APIs that work with any language. The platform is language-agnostic - you configure assistants via JSON and handle webhooks in your preferred backend technology.
Vapi primarily provides phone numbers for US and Canada. International deployments require external telephony providers with SIP integration. WebRTC calls work globally. The underlying STT/LLM/TTS providers support 20+ languages, but telephony coverage varies by region.
See Vapi plans and find the right tier for your needs.
See Pricing Plans →Still not sure? Read our full verdict →
Last verified March 2026