$0.05 per minute. No fine print.

The voice engine: orchestration (turn-taking, interruption handling, tools, knowledge base retrieval) plus Telnyx-hosted speech-to-text (including Deepgram models) and Telnyx-hosted text-to-speech (Telnyx Ultra, Natural, NaturalHD, Qwen3TTS, Inworld, Rime, Resemble, Murf). It is the same layer Retell prices at $0.055/min before add-ons.

LLM usage is billed per token at the rates in the tables above, not per minute. Models on Telnyx GPUs are billed at priority-tier inference rates (the Priority tab on /pricing/inference-api). Bundling into the minute rate would overcharge short prompts and cap long ones. On Telnyx-owned GPUs, Kimi typically works out to ~$0.006 per minute for a realistic production agent, with prompt caching and no prompt-size multipliers. Managed frontier models from Anthropic and OpenAI are billed at provider list rates on the same Telnyx invoice.

Telephony is billed at carrier cost on Telnyx's own network: SIP trunking from $0.0032/min inbound and $0.005/min outbound in the US, plus a $0.002/min Voice API platform fee. US local numbers are $1.00/mo with volume pricing from $0.79/mo. Everything lands on the same invoice as the voice engine, with no separate telephony vendor.

On connected call duration only. Each call is billed in 60-second increments: the connected duration of each call is rounded up to the next full minute, applied per call. For example, a 15-second connected call is billed as 60 seconds; a 90-second connected call is billed as 120 seconds. Pay as you go has no monthly minimum; Growth and Enterprise carry one. On every plan you are not charged for ringing or failed attempts.

What's inside the sticker. Vapi's $0.05 covers orchestration only: speech-to-text, text-to-speech, the LLM, and telephony are all passed through on top, which is how a production stack ends up at $0.12–$0.42 per minute. Telnyx's $0.05 includes orchestration, hosted STT, and hosted TTS; add LLM tokens at GPU rates and telephony at carrier cost and a realistic agent lands near $0.06 per minute, with no monthly platform fee.

Most teams migrate incrementally. Bring your prompts and tool definitions into the Telnyx voice AI agent builder, port your US and Canadian numbers in free, and route traffic number by number: you can run both platforms side by side during the cutover and shift volume as you validate call quality and cost.

Growth is a $2,000 monthly minimum spend commitment on your account, and its 15% discount applies to the Voice, Messaging and Numbers lines of your bill: the telephony under each call, the numbers you hold, and any SMS your agent sends. The voice-engine minute rate and LLM token rates are the same on every plan; that spend counts toward the minimum but is not discounted. So the test is the size of your Voice, Messaging and Numbers charges in dollars: once 15% off them outweighs committing to $2,000 a month, Growth wins, whatever share of the bill the engine and LLM make up. Automatic volume discounts and support 24/7 via chat or call come with it. Same platform, same models, same network on every plan.

Pay as you go includes 500 concurrent calls, 100 API requests per second, and 50 SMS per second. Enterprise removes them entirely on dedicated infrastructure with private network interconnect. If you need more headroom on any plan, talk to us.