TL;DR
- Vapi fits developers who want to choose providers and own the integration logic. Its $0.05 hosting minute sits alongside speech, model and carrier costs; bringing your own keys changes billing, not whether those services cost money.
- Retell is the practical default for a conventional qualification, CRM and booking workflow. Its connected tools reduce custom integration work, while its price still depends on the selected model, voice and telephony.
- Bland fits structured conversational pathways. Its AI rate bundles the model and speech services, with telephony separate. Enterprise is the relevant purchase for its documented warm-transfer feature.
- On the illustrative 4,000-minute workload below, selected communication and AI components cost about $486 on Vapi, $497 on Retell and $617 on Bland Start. Operating ownership can outweigh that difference.
Define the sales outcome
The buying decision starts with what must happen before the call ends. An inbound lead qualification agent may look up the caller, determine fit and route the opportunity. An appointment follow-up agent may confirm a slot and write a result. A more complex sales operation may need territory rules, account ownership and a human introduction.
All three platforms can participate in these workflows. The distinction is where the business logic lives, how much the buyer configures and which features require a larger commercial package. A pleasant conversation without the right CRM record or booked event is unfinished work.
This guide compares the voice platforms on the same answered-call workload. It does not treat a vendor’s demo, claimed latency or published case study as a measured performance result for your sales process. The recommendation follows documented capabilities and explicit economic assumptions.
The platform comparison
| Dimension | Vapi | Retell | Bland |
|---|---|---|---|
| Build approach | Assistant configuration and developer-owned tools | Prompt agents, connected provider tools and custom functions | Conversational pathways with live webhook nodes |
| Cost architecture | Hosting plus chosen providers and carrier | Voice infrastructure, model, voice and telephony components | Bundled AI minute plus carrier; package fee where applicable |
| Integration advantage | Custom backend logic and provider choice | Connected CRM/calendar actions alongside custom endpoints | Explicit branches, extracted variables and API responses |
| Human handoff | Transfer Call tools; configure transport and destination | Cold, warm and agentic warm transfer options | Standard transfers; documented warm-transfer feature requires Enterprise |
| Best fit | A product or engineering team owning the voice application | A practical CRM-aware qualification or booking service | Structured flows, especially an Enterprise implementation with handoff requirements |
| Poor fit | A team expecting provider choice to remove implementation work | A buyer expecting the lowest advertised minute to cover every configuration | A buyer budgeting Start while requiring Enterprise warm transfer |
Sources: Vapi Function tools, Vapi Transfer Call, Retell integration tools, Retell transfer options, Bland live webhooks, Bland warm transfer.
Vapi
Vapi is attractive when the voice agent is part of a software product or a company-owned application. A Function tool has a defined name and parameters and connects to server-side logic. That makes custom lead routing, account lookups and booking rules natural places to extend the agent. It also creates an integration that someone must maintain. Vapi Function tools.
Provider choice is useful when a team has a preferred speech provider, needs a particular voice or wants to adjust model cost independently. Bringing provider keys means those services can bill the buyer directly. It does not make their consumption free; the complete economic view includes those invoices. Vapi provider keys.
Current new-customer packages separate usage from support and capacity. Usage-only includes four concurrent calls; Core expands that to ten and adds a support response commitment. The package supplements usage billing. It does not replace the hosting and provider charges. Vapi package documentation.
Choose Vapi when that control has a concrete purpose. It is a less convincing choice for a small operations team that would have to recruit an engineer merely to connect a conventional calendar and CRM.
Retell
Retell is the strongest starting point here for a fairly standard qualification and booking service. Prompt agents can use connected CRM, helpdesk and calendar tools, with the provider connection and tool responses exposed in the agent workflow. A custom function remains available for business logic that does not fit a connected action. This is a more useful description than saying every integration needs a bespoke server. Retell integration tools.
For sales, the useful flow is a lookup before qualification, followed by the correct routing or booking action. A CRM action during the conversation can influence the next step. A post-call summary sent later cannot do that job.
Retell’s transfer tool supports cold, warm and agentic warm modes. The human handoff still needs an appropriate destination and fallback when the rep is unavailable. A caller who reaches voicemail should become a callback outcome rather than a successfully connected opportunity. Retell transfer documentation.
Its public pricing provides both a range and component rates. Use the selected combination rather than treat $0.07 as a universal deployed price. The displayed $0.11 example contains $0.055 infrastructure, $0.04 model and $0.015 voice, with zero telephony selected. Retell pricing.
Bland
Bland’s pathways suit a process whose branches should be visible: identify the customer, ask the relevant questions, retrieve a record and choose the next action. A webhook node makes an API request during the live call and can use extracted information. That is a practical fit for a structured qualification flow; the backend still owns the business transaction. Bland webhook nodes.
Its AI rate includes the language model, transcription and generated speech. Telephony is separate. Start has no platform fee and a $0.14 AI minute; Build costs $299 monthly with a $0.12 AI minute. Enterprise is custom-priced. Bring-your-own-telephony customers handle their carrier directly and do not pay Bland transfer fees under the current published terms. Bland pricing.
Warm transfer is a consequential plan gate. Bland documents a second call that briefs the rep before the calls merge, with timeout and fallback handling. That feature requires Enterprise. The Start budget in this guide therefore does not represent a warm-transfer deployment. Bland warm-transfer documentation.
Choose Bland for the pathway approach or a supported Enterprise rollout. Do not select its cheapest tier and assume it includes the enterprise handoff and dedicated scheduling features.
One call scenario and four cost layers
Assume 1,000 answered US outbound calls averaging four AI minutes, giving 4,000 AI minutes per month. Spread calls across 20 working days at 50 calls daily. The campaign records an outcome and arranges later follow-up; there are no live transfers in this baseline.
Use one existing Twilio US local number and an outbound carrier component of $0.014/minute. The number is $1.15 monthly, giving $57.15 for the selected carrier and number components. This is a common external carrier budget, rather than three different included-telephony promises. Twilio US voice pricing.
For Vapi, assume transcription at $0.01, model usage at $0.02 and voice at $0.02 per minute. Those are editorial configuration assumptions, not fixed rates for every provider. Add the published $0.05 hosting minute and $29 Core package. For Retell, use its displayed $0.11 AI combination with custom telephony. For Bland, use the published AI rates with bring-your-own telephony.
| Selected component | Vapi with Core | Retell example | Bland Start | Bland Build |
|---|---|---|---|---|
| Platform/voice infrastructure | 4,000 × $0.05 = $200 | 4,000 × $0.055 = $220 | Included in AI rate | Included in AI rate |
| Transcription assumption | $40 | Included in selected voice infrastructure | Included | Included |
| Model | Assumed $80 | 4,000 × $0.04 = $160 | Included | Included |
| Voice | Assumed $80 | 4,000 × $0.015 = $60 | Included | Included |
| Bundled AI usage | , | , | 4,000 × $0.14 = $560 | 4,000 × $0.12 = $480 |
| Package fee | $29 | $0 in this pay-as-you-go example | $0 | $299 |
| Common carrier and number | $57.15 | $57.15 | $57.15 | $57.15 |
| Selected monthly subtotal | $486.15 | $497.15 | $617.15 | $836.15 |
Pricing sources: Vapi, Retell, Bland.
Exclude implementation, taxes, recording/storage, additional numbers, failed-attempt billing, premium features and transfer/conference legs. Retell’s optional Batch Call charge is $0.005 per dial: 1,000 batch-launched dials would add $5, and additional attempts increase that count. The table is a comparable component model, not an all-in invoice forecast. Retell component pricing.
Bland Build’s rate saving alone offsets its $299 fee at 14,950 AI minutes monthly: $299 ÷ ($0.14 − $0.12). Capacity can justify Build before that point. Start publishes ten concurrent calls and 100 calls daily; Build publishes 50 concurrent calls and 2,000 daily. Our evenly spread 50-call day fits Start’s advertised daily cap, while a concentrated campaign could require another package. Bland pricing.
Capacity and operating economics
Four thousand monthly minutes says little about peak concurrency. Calls concentrated into a launch window need more simultaneous capacity than the same calls spread across the month. In our illustration, cap active conversations at ten. Vapi Core supplies ten; Retell’s public pay-as-you-go plan includes 20; Bland Start supplies ten. Vapi package limits, Retell pricing, Bland pricing.
Now assume 250 qualified follow-ups. Before operating labor, the selected subtotals equal $1.94 per qualified follow-up on Vapi, $1.99 on Retell and $2.47 on Bland Start. The equal outcome assumption isolates cost; it is not a conversion-rate benchmark.
Suppose Vapi requires eight maintenance hours monthly while the conventional Retell or Bland flow requires four, at an assumed engineering cost of $75/hour. The modeled operating subtotals become $1,086.15, $797.15 and $917.15 respectively. That reverses the small Vapi-versus-Retell component difference.
For a team already maintaining the Vapi integration, assume the same four-hour maintenance burden. Vapi then becomes $786.15, only $11 below Retell. Migration would need a workflow or reliability benefit to justify itself; shaving that amount from the usage bill is a weak reason.
A model change can also dominate the carrier difference. Adding $0.04 per AI minute costs another $160 on this workload. The right model is the least expensive one that completes the actual qualification and tool steps reliably, rather than the lowest unit-price model regardless of outcome.
Integrations that finish the job
Consider an existing customer calling about a new requirement. The agent should retrieve the correct account, route to the current owner and avoid creating a duplicate lead. A new prospect should receive the relevant qualification questions, then a committed next step with an owner and time. These are different paths even if the spoken introduction is similar.
Vapi favors custom tool logic when account rules are company-specific. Retell’s connected tools can cover conventional provider actions, with custom functions for exceptions. Bland’s pathway webhook can make the lookup and use the response in the next branch. None of those descriptions guarantees the CRM mapping is correct automatically.
For bookings, record the event identifier after the calendar accepts the transaction. A spoken preference for Tuesday afternoon is not a booked appointment. For a failed write, retain a callback task and tell the caller what actually happened. For live transfers, record whether a human answered rather than count the transfer attempt as success.
Use the signal-based outbound playbook to decide who should enter the campaign and why. More automated call capacity does not improve an irrelevant list.
Voice quality and response time
A headline response-time claim is rarely measured on the same carrier, model, prompt and tool workload as another vendor’s. This guide therefore makes no latency winner claim. The useful sales experience includes interruption handling, accurate names and dates, short tool waits and a clear recovery when an action fails.
Prioritize the slowest consequential moment. A fast greeting followed by an eight-second CRM lookup can still feel poor; a slightly slower initial response with accurate routing may produce the better completed interaction. Keep qualification questions brief and avoid lengthy narration while a backend action is pending.
That judgment supports Vapi when provider control serves a known requirement, Retell when the connected workflow reduces implementation burden, and Bland when explicit pathway branches suit the process. Voice preference alone should not obscure an unavailable handoff or an unfinished calendar action.
Which platform to choose
Choose Vapi for an engineering-owned voice product, custom routing rules or a deliberate provider strategy. Its cost visibility is useful when someone owns both the configuration and the external invoices.
Choose Retell as the default starting point for a conventional CRM-aware qualification and booking service. Its connected tools and transfer options provide the strongest practical match here without forcing the purchase into a developer-only description.
Choose Bland for structured pathways and a deliberate supported deployment. Start can suit a bounded follow-up campaign; Enterprise is the relevant basis when the process requires the documented warm handoff. Use the contract’s configured fee and rate rather than substitute Start pricing for that deployment.
For the workload above, Retell is the recommendation for a new conventional service. An existing competent Vapi team should usually keep its working integration. A Bland buyer should choose on pathway and implementation fit, with capacity and feature gates included in the decision.
FAQ
Does bringing your own Vapi keys make provider costs zero?
It can move provider charges to your own accounts. Those invoices still belong in the call budget alongside Vapi hosting.
Is Retell’s $0.11 example an all-in call price?
The displayed combination has zero telephony selected. The carrier, optional features and chosen configuration can add cost.
Does Bland bundle telephony into its AI minute?
Its current commercial page bundles the model and speech services but bills telephony separately. Bring-your-own telephony also changes its transfer-fee basis.
Is Bland Build always cheaper than Start?
Its lower minute rate offsets the $299 base at 14,950 monthly AI minutes on rate savings alone. Higher capacity can justify it sooner, while Enterprise features remain a separate purchase.
