Vapi Hidden Costs: What Their Pricing Page Doesn't Tell You
Published March 2026 · 10 min read
Vapi's pricing page says "$0.05/min." That's the platform fee. It does not include the LLM, speech-to-text, text-to-speech, or telephony costs that make up the majority of your actual per-minute spend. When you add it all up, most Vapi users pay $0.14-0.25 per minute.
This isn't a hit piece — Vapi is a capable platform. But their pricing structure is confusing, and many teams don't realize the true cost until they see their first invoice. Here's the full breakdown.
The real cost breakdown
Estimated costs for a typical Vapi deployment using GPT-4o, Deepgram Nova-2, and ElevenLabs Turbo v2:
| Component | Provider | Est. cost/min |
|---|---|---|
| Platform fee | Vapi | $0.05 |
| LLM inference | OpenAI GPT-4o | $0.03-0.08 |
| Speech-to-text | Deepgram Nova-2 | $0.015-0.025 |
| Text-to-speech | ElevenLabs Turbo v2 | $0.03-0.06 |
| Telephony | Twilio / Vonage | $0.01-0.02 |
| Total estimated cost | $0.14-0.25 | |
Estimates based on publicly available pricing from each provider as of March 2026. Actual costs depend on volume, negotiated rates, and usage patterns.
Vapi vs. Vociply: total cost comparison
| Monthly volume | Vapi (est. all-in) | Vociply (all-in) | Difference |
|---|---|---|---|
| 1,000 min | $140-250 | $80-120 | Save 40-50% |
| 10,000 min | $1,400-2,500 | $700-1,000 | Save 50-60% |
| 100,000 min | $14,000-25,000 | $5,500-8,000 | Save 60-70% |
Vociply pricing includes platform, STT, LLM, and TTS. Telephony at carrier rates. Volume discounts available for Enterprise.
Why stacked pricing hurts at scale
Unpredictable invoices
With 4-5 separate cost components, your invoice varies based on call duration, LLM token usage per call, and voice provider rates. Hard to forecast and harder to budget for.
Compounding at volume
At 100K+ minutes/month, even $0.05/minute in hidden costs adds up to $5,000/month. That's $60K/year that's not visible on the pricing page.
Vendor lock-in through complexity
Managing 4-5 vendor relationships for one product creates switching costs. If you want to change your TTS provider, you're changing your Vapi config, not just swapping an API key.
Where Vapi wins
In fairness, Vapi's stacked model has advantages for teams that want maximum control:
- • BYO keys: Bring your own OpenAI, Deepgram, and ElevenLabs keys. If you have negotiated enterprise rates, this can reduce costs.
- • Model flexibility: Mix and match LLMs, STT, and TTS providers per agent. Useful for teams that want to experiment.
- • Developer ecosystem: Strong SDK, active community, and more third-party integrations than most competitors.