Home › Solutions › Voice AI: Cost per Minute
Use caseVoice AI: Cost per Minute
A voice agent pays three times per minute: to transcribe, to think and to speak. The LLM part grows with every turn because the conversation is resent.
Defaults use GPT-6.1 Sol's list prices ($2 input, $0.10 cached, $10 output per 1M tokens, checked October 30, 2026). Enter your own model's prices. Speech prices are examples; enter your providers' rates.
Three cost layers
- Speech-to-text, usually priced per minute of audio.
- The LLM, priced per token, called on each turn with the growing conversation.
- Text-to-speech, priced per character or minute.
How to cut it
- Cache the system prompt and tool definitions, which repeat on every turn.
- Keep responses short: they cost output tokens and speech time.
- Use a smaller model for simple turns.
- Summarize long calls instead of resending everything.
Frequently asked questions
How much does voice AI cost per minute?
Speech-to-text plus text-to-speech plus LLM tokens for every turn in that minute. Enter your providers' prices in the calculator.
What's the biggest cost in a voice agent?
It depends on your providers, but the LLM part grows with conversation length because history is resent on every turn.
Want us to find your 20%?
Book a free 15-minute API cost teardown. Bring last month's provider usage and we'll show you your biggest waste and the first fix we'd ship.
Get a Free 15-Minute API Cost Teardown