open Voice AI economics

Voice AI costs, decoded.

Build a real per-minute estimate across platform, speech, intelligence, carrier, and recording. Every number is inspectable. Every source is public.

live_cost_trace.log
10Kminutes / month
5.0avg call minutes
3stacks compared

What 10,000 outbound minutes can cost

Same Deepgram, GPT-5 mini, integrated voice, and Twilio outbound assumptions. Bundled platforms include their speech and model layers.

Vapi$958.33$0.0958/connected min
Retell AI$1,008.33$0.1008/connected min
Bland Start$1,540$0.1540/connected min
Data statuscurrentverified 2026-07-18

Plan the whole call, not one line item

Useful calculators for the decisions that happen before a Voice AI launch: stack selection, carrier spend, burst capacity, and business case.

Published rates, with context

Advertised per-minute rates are rarely equivalent. These profiles explain whether the price includes infrastructure, STT, TTS, LLMs, telephony, or none of them.

Compare the choices people actually shortlist

Each page explains the pricing structure, ownership trade-off, and a like-for-like 10,000-minute scenario.

Economics now. Open call infrastructure next.

Voice OSS is created by Himanshu Maurya (codingstark), who is also working toward an open-source Voice AI calling stack designed around Cloudflare's edge platform.

creator

Himanshu Maurya (codingstark)

Builder of OSS.codes and practical open-source developer tools.

active direction

Realtime media, Workers, Durable Objects, WebSockets, and open agent adapters on a Cloudflare-based stack. This site is the research and cost-modeling layer for that work.