Realtime voice agents, compared.
Packaged platforms advertise a low platform fee - then bill the model, the voice, and telephony on top. Here's the all-in per-minute next to what each one advertises, plus where Omni's one-socket, telephony-native agent fits. One flat rate, no token math, no asterisk.
All-in price per minute (USD)
Lower is better. PyAI in teal. The bar spans each platform’s typical all-in range.
All-in, billed per second, speech + brain + telephony in one rate. $0.05/min Omni API, $0.08/min for the no-code Agents feature.
Advertised from ~$0.09/min self-serve → all-in $0.11-$0.14/min once LLM, voice, and telephony are added.
Advertised agent minutes + LLM → all-in $0.10-$0.15/min once LLM, voice, and telephony are added.
Advertised no-code plans → all-in $0.11-$0.16/min once LLM, voice, and telephony are added.
Advertised ~$0.05/min platform fee → all-in $0.10-$0.31/min once LLM, voice, and telephony are added.
Advertised ~$0.07/min platform → all-in $0.13-$0.31/min once LLM, voice, and telephony are added.
PyAI from our rate card (all-in, billed per second). Competitor figures are advertised pricing plus STT/LLM/TTS/telephony passthrough estimates as of June 2026 - composite and config-dependent, so verify before relying on them.
Why teams pick Omni
- ~390 ms median voice-to-voice, in-region, with barge-in - human conversational pace
- One speech-to-speech model, built for 8 kHz phone audio
- One WebSocket - no stitching, no cross-vendor hops
- Grounded in your knowledge base, calls your tools
- OpenAI-realtime-compatible alias to drop in
- $0.05/min all-in - speech, brain, and telephony in one bill
The numbers
| Provider | Advertised | All-in / min | Source |
|---|---|---|---|
| PyAI OmniPyAI | No platform fee | $0.05-$0.08 | PyAI rate card |
| Bland AI | from ~$0.09/min self-serve | $0.11-$0.14 | bland.ai (advertised), all-in estimated |
| ElevenLabs Agents | agent minutes + LLM | $0.10-$0.15 | elevenlabs.io (advertised) + LLM passthrough est. |
| Synthflow | no-code plans | $0.11-$0.16 | synthflow.ai (advertised), all-in estimated |
| Vapi | ~$0.05/min platform fee | $0.10-$0.31 | vapi.ai (advertised) + STT/LLM/TTS/telephony passthrough est. |
| Retell AI | ~$0.07/min platform | $0.13-$0.31 | retellai.com (advertised) + passthrough est. |
Methodology: all-in cost per minute of conversation (speech + brain + telephony). PyAI Omni is one flat $0.05/min rate with everything included - lower than any packaged platform here. Packaged figures are the advertised platform fee plus passthrough estimates as of June 2026; confirm with each provider.
What about the raw model APIs?
Gemini Live, Amazon Nova Sonic, and OpenAI Realtime undercut everyone on token price - but they're a raw model, not a product. No managed telephony, no orchestration, no agent. You're back to building the Frankencascade around them.
| Raw model API | Token price | What you still build |
|---|---|---|
| Amazon Nova Sonic | ~1.2-1.5¢/min | Raw S2S model, no telephony, no orchestration. |
| Gemini Live | ~1.4-3¢/min | Raw S2S model, you still build the agent. |
| OpenAI Realtime (cached) | ~5-10¢/min | Token/audio billed; production calls run higher. Not telephony-managed. |
Token prices are public estimates as of June 2026and aren’t an all-in comparison - they price the model, not a telephony-managed agent. We list them for honesty, not to claim a price crown.
See Omni run a call, end to end
Acme Dental’s front-desk agent - every step powered by PyAI.
Hear <-> Speak repeats every turn, with barge-in - then the call becomes data.
Each step is a PyAI model. Press play to watch them hand off - or tap any stage to jump.
- Hear
pyai-hear - Omni brain
pyai-omni-realtime - Speak
pyai-voice
Transcript, summary, intent, and disposition - emitted as JSON when the call ends.
One model. One hop. Built for the phone.
Start free with $50 in free credits - no card. Bring a phone number and ship a voice agent today, all-in at $0.05/min.
Agents is live in beta. Sandbox keys have daily limits and never touch billing.