Voice AI Agents
$0.07 per minute. One pipeline.
Orchestration, speech recognition and speech synthesis in one rate. Standard models included. Premium LLM tokens and telephony on top, from the same prepaid credit.
How a minute is built
Three charges. One invoice.
$0.07/min
Agent runtime, STT, included TTS voices, interruption handling. Standard models ride free on this rate.
Included or per token
Economical models are covered by the engine rate. Frontier models bill input and output tokens with our markup.
From $0.003/min
Carrier cost to the destination, by country. Browser sessions use a flat WebRTC leg instead.
Voice engine
What $0.07 covers
The managed pipeline every voice agent runs on. Billed per minute of connected time, from answer to hangup.
$0.070/min
Ledger lines post at $0.063/min (6.25¢). The headline rounds up so the page matches the rest of pricing.
- Full agent runtime: turn-taking, tools, knowledge retrieval
- Speech-to-text in real time
- Text-to-speech on included voices
- Interruption handling and barge-in
- Transcripts on every call
- Standard LLM models (Qwen, Kimi) included
LLM models
Models you can put behind the agent
Standard models are bundled into the engine rate. Premium models add token charges, markup included in the rates below.
| Model | Tier | Input / 1M | Output / 1M | Est. / min |
|---|---|---|---|---|
| Qwen3 235B | Standard | Included | Included | — |
| Kimi K2.6 | Standard | Included | Included | — |
| Claude Haiku 4.5 | Premium | $1.25 | $6.25 | +$0.011/min |
| GPT-5.2 | Premium | $2.19 | $17.50 | +$0.022/min |
| GPT-4o | Premium | $3.13 | $12.50 | +$0.025/min |
Per-minute estimates assume ~6,000 prompt tokens and 500 output tokens per minute of talk time, a realistic production agent with tools and retrieval. Actual billing is per token for premium models.
Telephony
Carrying the call
The phone network is priced by destination, not by the agent. Same rates as the rest of the platform.
PSTN (phone)
From $0.0030/min
Landline and mobile rates by country. Full table on the voice rates page.
WebRTC (browser)
$0.0024/min
In-browser sessions have no carrier leg. This is the web calling charge only.
Phone number
Upfront + monthly
Purchase fee plus first month due today, then monthly only. Same line items as the dashboard buy wizard.
See number pricingEstimate
Your numbers, itemized
Set volume and model. Every line maps to what the ledger posts: engine, LLM, telephony.
Estimated all-in
$0.066/min
≈ $655.00 / month at 10,000 minutes
- Voice engine
- $0.063/min
- LLM · Qwen3 235B
- Included
- Telephony
- +$0.0030/min
Estimate only. Premium LLM uses a fixed token profile; real calls vary with prompt size and how much the agent speaks. Telephony uses our published floor. Your destinations may cost more.
FAQ
How billing works
The managed voice pipeline: agent orchestration (turn-taking, tools, knowledge retrieval), real-time speech-to-text, speech synthesis on included voices, interruption handling, and transcripts. Standard LLM models (Qwen, Kimi) are covered by this rate. Premium model tokens and telephony are billed separately.
Ship a voice agent on one credit.
No separate STT, TTS or LLM vendor accounts. Start free, pay as you go.