Retell per-minute voice economics

This pricing model makes Retell behave more like cloud infrastructure than classic SaaS. Revenue rises every time a customer sends more calls through the system, because each live minute bundles three separate cost centers, speech, reasoning, and phone connectivity, into one metered workflow. That creates strong expansion when customers move from a single use case, like appointment booking, into collections, claims, support, and overflow call handling.1 2 3

• Retell sits as an orchestration layer on top of outside providers. It adds a $0.055 per minute platform fee, then passes through model, speech, and voice costs from vendors like OpenAI, Anthropic, Deepgram, Cartesia, and ElevenLabs. That keeps setup flexible, but also ties margins to upstream pricing.1
• This is the same basic billing logic used by Vapi and other developer first voice platforms. Customers are not buying seats. They are paying for every minute the bot listens, thinks, speaks, and stays on the phone, which makes cost tracking look more like Twilio or AWS than Zendesk.2 3
• The upside is that usage can compound fast inside one account. A healthcare group or insurer that proves ROI on inbound scheduling can expand into reminders, intake, renewals, and after hours triage, pushing more minutes through the same stack without a full rip and replace of call infrastructure.1 2

The next step is moving from metered voice runtime into higher value software around routing, QA, testing, analytics, and vertical workflows. If Retell owns more of the full contact center workflow, it can keep per minute pricing as the entry point while capturing more revenue and more margin above the raw voice stack.1 2