Vapi pricing, and what the pass-throughs add

Start with what Vapi's own pricing page says, because it says this part plainly and we are not going to argue with it: “Vapi Hosting Cost Excludes Model Provider Costs.” Underneath that line it lists “Model Provider Cost (STT, LLM, TTS) … At cost ($0 if you bring your own API key).” The platform rate itself is $0.05 a minute — a real number, on the page, and lower than Komms's $0.08. Vapi also charges for calls running at the same time: a usage-only account includes 4 concurrent calls, its $29-a-month Core package includes 10, and each line past that is $10 a month.

Komms's own pricing table has one per-use row — call minutes — and no row for concurrency. Call minutes is $0.08 a minute with inbound and outbound the phone call, the AI and a standard voice already inside that rate, billed in 6-second increments; the only other charge is $15 a month per phone number. Anything you bring yourself, such as a premium voice provider, bills to you directly. Komms has no flat entry rate of its own to compare against a monthly platform fee — it is priced purely per minute, the same shape as Vapi's platform charge, just a higher number and with more already folded in.

Here is where this page has to stop rather than keep going the way the Retell and Bland pages did. Those two vendors publish either a range or a paired fee that resolves to a dollar figure on its own. Vapi does not: what a minute actually costs on Vapi is $0.05 plus whatever your speech-to-text, model and text-to-speech providers charge you directly, and that third piece is not a number Vapi publishes, not a number we have measured, and not a number we are going to invent so this page can print one final total. A buyer running a cheap, efficient model against a discounted STT and TTS stack could land the whole call meaningfully under $0.08. A buyer running a frontier model with premium voice synthesis could land well above it. Both of those buyers are reading the same $0.05 headline rate, and it means something different to each of them.

That is the honest shape of the comparison, and it is also the reason we are not going to pretend one exists that we can size for you. On raw platform rate, Vapi is cheaper than Komms — three cents a minute cheaper, before anything else is added. Whether the finished bill is cheaper depends on rates only the buyer has: whatever they already pay their STT vendor, their model provider and their TTS vendor, none of which we can see or publish on their behalf.

What does not depend on any of that is what each architecture asks you to run. A Vapi deployment is Vapi's platform plus three more vendor relationships — a speech-to-text account, a model account, a text-to-speech account — each with its own invoice, its own rate limits, its own outage page, and its own support queue to open when a call goes wrong. A Komms deployment is one: the $0.08 a minute you pay us, phone calls included — a second only if you bring your own premium voice. Count the carrier on Vapi’s side too, because a Vapi deployment needs one, and the tally is one against five. That is a count of vendors, not a price.

On platform rate, Vapi is $0.05 a minute and Komms is $0.08. Vapi is self-serve and developer-first. Speech-to-text, the model and the voice are separate accounts, so a team with its own provider contracts can plug them in and swap providers. Komms is self-serve too, with 50,000 free tokens. Its rate includes the model and a standard voice, and you can bring your own voice.

A Vapi deployment has four vendors that each charge per use — Vapi, speech-to-text, the model and text-to-speech — plus a carrier: five invoices and five support queues. Komms is one invoice, plus a premium voice provider if you bring one.