We believe in pricing so competitive that you won't find a lower price elsewhere, not even by building your own stack.
Spark at 2¢/min, vs a typical 10¢/min itemized stack.
Flat per-minute pricing. Move a workload up or down a tier whenever you want.
Somewhat slower latency, but acceptable for most use cases. High audio understanding, average instruction following.
Like Spark, but much faster. Slightly lower audio understanding, with higher instruction following.
Strong latency with the best audio understanding and instruction following. Combines thinking and non-thinking LLMs.
Fast — but it mishears and drifts off-script, exactly where Bolt and Storm pull ahead.
All tiers show base per-minute rates. Premium voices, long prompts, and demo numbers are optional add-ons, billed separately.
Every tier ships with the full stack.
Pay only for what your use case needs.
Competitors charge platform fees on top of model costs and start at 7¢/min — and go much higher than that. ThunderPhone keeps the full stack affordable and charges no platform fee. Cost per live call minute, reviewed May 2026.
Vapi shown as a best-achievable typical stack: its platform fee plus transcriber, model, voice, and telephony. Competitor totals vary by provider choices.
We think voice AI should work like AWS or Costco: low margins at high volumes, making customers successful at fair, low prices. We don't add a platform fee on every minute.
We'll match or beat the performance of any competing stack on your use case, and we'll beat the pricing as well. We price so low that you won't even save money if you build your own stack in-house. (Reach out if you think we're wrong!)
For high-volume use cases including but not limited to outbound with long hold times, voicemail, and screeners, we can reach 1¢/min or less with an enterprise agreement. Reach out to learn more.
See enterprise pricingPick a tier, write a prompt, connect your number.