A voice agent is a stack — speech recognition, a language model, a voice, and a phone line. We show you the real cost of every layer, at provider rates. What Yaapi charges on top is scoped to your build.
| Layer | Provider & model | Rate |
|---|---|---|
| Speech‑to‑text | Deepgram Nova‑2 | $0.0077 / min |
| Language modelyou choose per build | OpenAI GPT‑4o mini | $0.15 / $0.60 · 1M tok |
| DeepSeek Chat | $0.27 / $1.10 · 1M tok | |
| Groq Llama 3.3 70B | $0.59 / $0.79 · 1M tok | |
| Anthropic Claude Haiku 4.5 | $1.00 / $5.00 · 1M tok | |
| Text‑to‑speech | ElevenLabs Flash v2.5 | $0.055 / 1k chars |
| ElevenLabs Multilingual v2 | $0.11 / 1k chars | |
| Telephony | Inbound phone minutes | $0.01 / min |
Token rates show input / output. Provider rates are set by each vendor and change over time — these are the rates Yaapi currently meters every call against (as of August 2026). You pick the model that fits your quality and budget.
The rates above are the underlying provider costs. Yaapi's own pricing is scoped to your build — every agent is custom-engineered for your business logic, not sold as a shrink-wrapped plan. Tell us what you need and we'll put a number to it.