Vapi promotes a deceptively simple promise: $0.05 per minute for voice AI calls.<\/a> Scan the pricing page for three seconds and you might think you\u2019ve found a bargain. But seasoned agencies know the $0.05 figure is only the orchestration fee \u2014 the tip of a much deeper cost iceberg<\/strong>.<\/p>\n\n\n\n The advertised price is almost never what you pay. Vapi says $0.05 per minute. That is true if you ignore the LLM, the transcription engine, the voice synthesis, and the telephony carrier. Add those in and you are at $0.15 to $0.31 per minute before your first call ends.<\/strong><\/p>\n\n\n\n This guide is a line\u2011by\u2011line dissection of what a single minute of Vapi voice AI actually costs. You\u2019ll get the real\u2011world figures agencies report after dozens of productions, learn why Vapi\u2019s dashboard often hides the true cost, and walk away with a clear model for pricing your services \u2014 and keeping your margins intact.<\/p>\n\n\n\n A Vapi voice call isn\u2019t one service. It\u2019s five distinct billable components<\/strong> stacked on top of each other, each invoiced by a different provider:<\/p>\n\n\n\n Real\u2011world all\u2011in cost: $0.15 \u2013 $0.36 per minute<\/strong> (commonly $0.23 \u2013 $0.33), depending on your model and provider choices.<\/p>\n\n\n\n This is the only fixed component. Vapi charges $0.05 per minute for hosting and orchestrating the call \u2014 managing the WebSocket, keeping latency low, and stitching together the other four providers. Every other cost is passed through from third\u2011party services you select.<\/p>\n\n\n\n The language model you pick is the biggest swing factor. GPT\u20114o can cost $0.08\u2013$0.20 per minute depending on prompt length and token usage, while smaller models like GPT\u20114o mini, Claude 3.5 Haiku, or Gemini Flash can bring this below $0.05\/min. Longer prompts and multi\u2011turn conversations with large context windows push this number higher, sometimes dramatically.<\/p>\n\n\n\n Voice quality directly affects this line. Premium providers like ElevenLabs with the most natural human voices run ~$0.036\/min, while budget options like Deepgram or Azure TTS cost ~$0.011\/min. Open source voices can lower this to $0.002\/min if you self\u2011host, but at the cost of setup complexity.<\/p>\n\n\n\n Deepgram (~$0.01\/min) is the default, but agencies often switch to Assembly AI ($0.00025\/min) or ElevenLabs ($0.00667\/min) based on accuracy needs. This is typically the smallest line item.<\/p>\n\n\n\n Vapi doesn\u2019t sell phone numbers. You bring your own telephony provider \u2014 Twilio, Vonage, or Telnyx \u2014 and pay that provider directly. Twilio inbound is ~$0.008\u2013$0.014\/min, outbound slightly more. Telnyx is often cheaper at $0.0055\/min.<\/p>\n\n\n\n Even when you understand the four\u2011layer model, Vapi\u2019s dashboard often obscures the real cost<\/strong>. Community forums are full of users reporting the same surprise:<\/p>\n\n\n\n “I can see the latency but I don’t see the costs.”<\/em><\/p>\n\n\n\n “The dashboard indicated our agent cost was approximately $0.09\/min, which is what we relied on when scaling volume. Actual charges came in at roughly $0.16\/min.”<\/em><\/p>\n\n\n\n “I use the GPT\u20114o cluster model for my assistant (which is $0.08\/min). But on the assistant dashboard, the average cost is ~$0.16\/min – why?”<\/em><\/p>\n\n\n\n The discrepancy comes from two places. First, the dashboard sometimes displays only the Vapi platform fee<\/strong>, excluding provider costs that appear later on separate invoices. Second, Structured Output and Analysis fees<\/strong>, charges for post\u2011call data extraction, are often missing from the live cost estimate but appear on the final bill.<\/p>\n\n\n\n For agencies, this is poison. How do you price a client engagement when you can\u2019t even see your own costs in real time?<\/p>\n\n\n\n Vapi offers a pay\u2011as\u2011you\u2011go Ad\u2011Hoc plan and several bundled tiers with included minutes. Here\u2019s the effective per\u2011minute breakdown after accounting for typical provider costs:<\/p>\n\n\n\n For enterprise deployments ($40,000\u2013$70,000+\/year contracts), Vapi bundles LLM, TTS, STT, and telephony into a single predictable invoice, but that\u2019s only available at scale.<\/p>\n\n\n\n Agency owners who\u2019ve deployed 40+ production agents have a consistent message: budget $0.23\u2013$0.33\/min all\u2011in, and don\u2019t trust the dashboard alone<\/strong>:<\/p>\n\n\n\n “A single 3\u2011minute call can cost $0.30\u20130.60. At 500 calls\/month, you’re looking at $150\u2013300\/mo just in usage \u2014 before you’ve paid a developer to build and maintain it.”<\/em><\/p>\n<\/blockquote>\n\n\n\n Vapi\u2019s $0.05\/min platform fee is the lowest among major voice AI platforms, but the total all\u2011in cost depends on your stack choices:<\/p>\n\n\n\n Vapi is the cheapest platform fee<\/em> per minute, but because you bring your own providers (BYOK), you carry the full cost of LLM, TTS, and STT yourself. The BYOK costs (ElevenLabs, OpenAI, Deepgram) are identical across platforms because they\u2019re the same providers.<\/strong> The difference is that platforms like Retell and Bland negotiate volume discounts with these providers and pass some savings on to you.<\/p>\n\n\n\n If you need maximum control<\/strong> and have the engineering bandwidth to optimize your stack, Vapi is unmatched. If you need predictable billing and faster deployment<\/strong>, Retell or Bland may be better fits. The tradeoff is always control versus predictability.<\/p>\n\n\n\n Given these costs, how do you charge clients and maintain margin?<\/p>\n\n\n\n Model 1: Cost\u2011Plus Markup.<\/strong> Track actual Vapi + provider costs per client, apply a 1.3\u00d7\u20131.5\u00d7 markup, and bill monthly. Risk: if costs spike (e.g., a client uses longer prompts), your margin compresses unless you monitor constantly.<\/em><\/p>\n\n\n\n Model 2: Flat Per\u2011Minute Rate.<\/strong> Charge $0.40\u2013$0.55\/min regardless of your underlying cost. Pro: predictable for both sides. Con: you need to know your average cost well. At $0.30\/min cost, $0.40\/min leaves you a 25% margin \u2014 slim.<\/em><\/p>\n\n\n\n Model 3: Tiered Subscription.<\/strong> $299\/month for up to 1,000 minutes, then $0.35\/min overage. Best for agencies with stable call volumes. Aligns client expectations with your cost structure.<\/em><\/p>\n\n\n\n Model 4: Flat Monthly Retainer.<\/strong> $400\u2013$1,000\/month based on forecasted volume. The Reddit thread from an agency owner captures the real market: “When I told him it would be 20 cents\/min, he literally said ‘no way’ \u2014 but he was willing to pay $400\/month fixed.”<\/em> The agency owner was worried about losing money \u2014 but at $400\/month for 1,000 minutes, that\u2019s $0.40\/min, more than enough to cover the $0.23\/min all\u2011in cost.<\/p>\n\n\n\n The pricing model you choose should reflect your client\u2019s psychology (some prefer fixed, others prefer usage\u2011based) and your ability to track real costs.<\/p>\n\n\n\n If you\u2019re relying on Vapi\u2019s native dashboard for cost tracking, you\u2019re flying partially blind. The Five white\u2011label Vapi portal pulls per\u2011client analytics directly from Vapi\u2019s API, calculates total cost (platform + providers), and applies your chosen pricing model automatically.<\/p>\n\n\n\n Vapi is an extraordinary platform for building voice AI agents. But its pricing model requires active management. Understand the five cost layers. Pick your providers based on both quality and cost. Model your pricing so your margins survive a spike in token usage. And if you\u2019re serving multiple clients, put a system in place that tracks costs automatically \u2014 because spreadsheets don\u2019t scale.<\/p>\n\n\n\n The agencies that win in voice AI aren\u2019t the ones with the best prompts. They\u2019re the ones who\u2019ve mastered the economics.<\/p>\n\n\n\n
\n\n\nTable of Contents<\/h2><\/div>
\n\n\n\nThe Real Cost Stack: Why $0.05 Is Just the Start<\/h2>\n\n\n\n
Component<\/th> What it does<\/th> Typical cost per minute<\/th><\/tr><\/thead> Vapi Orchestration<\/strong><\/td> Manages WebSocket connections, audio streaming, and provider coordination<\/td> $0.05<\/td><\/tr> LLM (Large Language Model)<\/strong><\/td> Powers the agent\u2019s conversation intelligence<\/td> $0.02 \u2013 $0.20<\/td><\/tr> TTS (Text-to-Speech)<\/strong><\/td> Synthesizes the agent\u2019s spoken responses<\/td> $0.01 \u2013 $0.07<\/td><\/tr> STT (Speech-to-Text)<\/strong><\/td> Transcribes what the caller says<\/td> $0.004 \u2013 $0.02<\/td><\/tr> Telephony<\/strong><\/td> Routes the actual phone call<\/td> $0.005 \u2013 $0.03<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n 1. Vapi Platform Fee (~$0.05\/min)<\/h3>\n\n\n\n
2. LLM Costs ($0.02 \u2013 $0.20\/min)<\/h3>\n\n\n\n
3. Text-to-Speech ($0.01 \u2013 $0.07\/min)<\/h3>\n\n\n\n
4. Speech-to-Text ($0.004 \u2013 $0.02\/min)<\/h3>\n\n\n\n
5. Telephony ($0.005 \u2013 $0.03\/min)<\/h3>\n\n\n\n
\n\n\n\nWhat the Dashboards Won’t Tell You<\/h2>\n\n\n\n
\n\n\n\nVapi Pricing Tiers: What Each Plan Actually Costs<\/h2>\n\n\n\n
Tier<\/th> Monthly fee<\/th> Minutes included<\/th> Effective per\u2011min (incl. providers)<\/th> Overage rate<\/th><\/tr><\/thead> Ad\u2011Hoc<\/strong><\/td> $0<\/td> Pay\u2011as\u2011you\u2011go<\/td> ~$0.18\/min<\/td> ~$0.18\/min<\/td><\/tr> Enterprise<\/strong><\/td> Custom<\/td> Custom<\/td> Negotiable<\/td> Negotiable<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\n\n\n\nWhat Real Agencies Are Reporting<\/h2>\n\n\n\n
\n
\n
\n\n\n\nVapi vs. Competitors: How Does It Stack Up?<\/h2>\n\n\n\n
Platform<\/th> Platform fee<\/th> Typical all\u2011in<\/th> Notes<\/th><\/tr><\/thead> Vapi<\/strong><\/td> $0.05\/min<\/td> $0.15\u2013$0.36\/min<\/td> Most flexible; you pick every provider<\/td><\/tr> Retell AI<\/strong><\/td> $0.07\u2013$0.23\/min<\/td> $0.15\u2013$0.30\/min<\/td> More inclusive pricing; fewer provider decisions<\/td><\/tr> Bland AI<\/strong><\/td> $0.09\u2013$0.12\/min all\u2011in<\/td> $0.09\u2013$0.12\/min<\/td> Most predictable; least flexible<\/td><\/tr> Safina<\/strong><\/td> $15\u2013$25\/month<\/td> ~$0.09\/min<\/td> Bundled telephony + AI; limited customization<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\n\n\n\nHow to Price Your Agency Services (Without Losing Money)<\/h2>\n\n\n\n
\n\n\n\nHow Five Closes the Visibility Gap<\/h2>\n\n\n\n
\n
\n\n\n\nTake Control of Your Vapi Economics<\/h2>\n\n\n\n
\n\n\n\nBuild Your Profitable Voice AI Agency With Five<\/h3>\n\n\n\n