Retell AI and Bland AI run voice agents on their own cloud, and neither is open source. Retell charges $0.055 a minute for voice infrastructure, then bills the model, voice and phone line you pick. Bland's single rate covers its own models, $0.14 or $0.12 a minute, plus phone. Retell signs a business associate agreement (BAA) free; Bland's needs Enterprise.
Key Takeaways
- Retell lets you pick the AI model; Bland runs only its own.
- Retell signs a BAA at no extra fee; Bland keeps it for Enterprise.
- Both sell on-prem only through Enterprise contracts, for closed-source code.
This post is part of our guide to Retell AI alternatives. The guide covers hosted and open-source options, and this post sets the two hosted platforms people most often weigh against each other side by side.
Retell AI vs Bland AI at a glance
Retell sells a platform you assemble from priced parts, while Bland sells one rate for a stack it builds itself. Each cell below says what you get, then the gap.
| Feature | Retell AI | Bland AI |
|---|---|---|
| How you pay | $0.055/min for voice infrastructure, then the model, voice and phone line, each per minute. But the total changes with every model and voice you pick | One per-minute rate that includes the AI models: $0.14 on Start, $0.12 on Build. But Build adds a $299 monthly fee, and phone time is billed on top |
| Free start | $10 in free credits. But it is a one-time credit, and pay-as-you-go calls stop when your prepaid balance reaches $0 | Start has no monthly fee and needs no card. But it caps you at 100 calls a day and 100 an hour |
| AI models | Dozens of OpenAI, Anthropic and Google models, OpenAI's realtime speech-to-speech models, or your own model server. But Retell bills the model itself, per minute | Bland's own AI models for the whole call, included in the rate. But you cannot pick a model provider on Start or Build |
| Builder | Single-prompt agents or a visual Conversation Flow. But knowledge bases add $0.005/min, and $8 a month each after the first 10 | Conversational Pathways, a visual node editor, plus Norm, which builds from plain-language instructions. But Norm is billed separately, by tokens |
| Concurrent calls | 20 included, then $8 a month each. But calls that burst past your limit pay $0.10/min extra on the whole call | 10 on Start, 50 on Build. But both plans also cap calls per day, at 100 and 2,000 |
| Phone numbers | $2 a month per number ($5 for US toll-free); your own numbers and SIP (Session Initiation Protocol) trunk at no charge. But Retell sells numbers only in the US and Canada | $15 a month per US or Canada number, or bring your own Twilio account. But SIP from other carriers is an Enterprise feature that Bland support has to switch on |
| Outbound calls | Batch calling at $0.005 a dial, and failed calls are not billed. But branded caller ID costs $0.10 per outbound call | Outbound runs at the plan rate. But every call attempt carries a $0.015 minimum, answered or not |
| HIPAA and security | SOC 2 Type 1 and Type 2, and a signed BAA at no extra fee. But removing personal details from transcripts is a paid add-on at $0.01/min | SOC 2 Type I and Type II, HIPAA, GDPR and PCI DSS. But the BAA and single sign-on are Enterprise only |
| EU data | GDPR covered by a data processing addendum. But Retell does not operate services inside the EU | Every part of a call can be pinned to one region, with EU data on EU servers. But data residency is Enterprise only |
| On-prem | On-prem offered through Enterprise sales. But the code is closed, with no published price or install details | Your own virtual private cloud (VPC), on-prem or fully offline setups on Enterprise. But the code is closed, with no published price |
Prices are list prices from Retell's pricing page and Bland's pricing page. Bland's billing page also lists a Scale plan, $499 a month at $0.11 a minute, which its pricing page does not show.
How each one bills a minute
Retell adds up separate lines for every minute, while Bland folds the AI models into one rate.
Retell's bill starts with a $0.055 per minute line called Retell Voice Infra. Retell lists no separate speech-to-text charge, so it appears to cover that too. The model and the voice each add a per-minute line, and so does the phone line if you use a Retell number. Retell's own example comes to $0.11 a minute for GPT 5.1 and one of its standard voices, before telephony, and calls that fail to connect cost nothing. The total moves with each choice, though: a larger model or an ElevenLabs voice raises it, and each add-on, from noise removal to AI quality checks, has its own charge.
Bland's rate covers every AI model in the call, with no token charges and no model bills passed through. Start costs $0.14 a minute with no monthly fee. Build costs $0.12 a minute plus a $299 monthly platform fee. Phone time sits on top, on your own carrier or on Bland's at cost, and each outbound attempt has a $0.015 minimum whether or not anyone picks up. For a team dialling long lists with low answer rates, that minimum adds up fast.
By our arithmetic from those list prices, 10,000 minutes a month comes to about $1,100 on Retell using its own $0.11 example, and $1,499 on Bland Build ($299 plus $0.12 a minute). Bland Start would come to $1,400, but at 100 calls a day, 10,000 minutes only fits if calls average more than three minutes. Phone time is extra on both. Our post Retell AI Pricing Explained prices every Retell line for three example businesses.
Choosing the AI model and building the agent
The biggest design difference is who picks the models. On Retell you do, and on Bland the vendor does.
Retell lists dozens of language models from OpenAI, Anthropic and Google, each with its own per-minute price, and it also sells OpenAI's realtime speech-to-speech models. You can connect your own model server instead. Agents are built as a single prompt or as a Conversation Flow, a visual editor where each node handles one part of the call. Retell bills the model itself, so switching to a cheaper model means picking another one from its list. Knowledge base use adds $0.005 a minute, and each knowledge base past the first 10 costs $8 a month.
Bland says it runs its own speech, language and voice models, built by one team, which is how one rate can cover them. Its builder, Conversational Pathways, is a visual node editor with nodes for webhooks, knowledge bases and transfers. Changes stay in a draft until you publish them, so live calls are not affected while you edit. Bland also offers Norm, which builds an agent from plain-language instructions. The trade is control. You cannot pick a model provider on Start or Build, and Norm is billed separately, by tokens.
Both builders draw a call as connected nodes, so a team used to one will recognise the other. Our comparison of Vapi vs Retell AI covers how Retell's Conversation Flow differs from building with Vapi's Squads.
Open Source Alternative to Vapi / Retell
Self-hosted voice agent platform — no per-minute fees
dograh-hq/dograh
Star on GitHub
Phone lines, HIPAA and where the data sits
Phone numbers, compliance and where calls are processed decide which deals you can sign. Retell includes more on its standard plan, while Bland keeps its compliance and private hosting options for Enterprise.
Retell AI. Retell sells US and Canada numbers only, at $2 a month ($5 for a US toll-free number), and charges nothing for your own numbers or SIP trunk. It reports SOC 2 Type 1 and Type 2. For healthcare, it signs a business associate agreement (BAA), the contract HIPAA (the US Health Insurance Portability and Accountability Act) requires before a vendor handles patient data, and Retell's compliance page says there is no extra fee to sign it. The same page says Retell does not currently operate services inside the European Union. Retell's security overview lists on-prem, which Retell offers through Enterprise sales. The platform is closed source, and Retell publishes no price or install details for on-prem.
Bland AI. Bland sells US and Canada numbers at $15 a month, accepts your own Twilio account, and connects other carriers over SIP on Enterprise, once Bland support switches it on. It reports SOC 2 Type I and Type II, HIPAA, GDPR (the EU's General Data Protection Regulation) and PCI DSS. Bland's trust page lists its own cloud, your VPC, on-prem GPU clusters and air-gapped setups, and Bland can pin every leg of a call to one region. All of that sits on Enterprise: the BAA, data residency, single sign-on and on-prem are not part of Start or Build.
What this means for your choice:
- Patient data without an Enterprise budget: Retell signs a BAA at no extra fee on pay-as-you-go. On Bland, the BAA needs an Enterprise contract.
- EU customers whose data must stay in the EU: Retell does not operate services inside the EU. Bland can pin every leg of a call to the EU, on Enterprise.
- Numbers outside the US and Canada, or your own carrier: Retell takes your own SIP trunk at no charge. Bland takes your own Twilio account, and other carriers over SIP only on Enterprise.
- The platform on your own servers: both offer on-prem only through an Enterprise contract, for closed-source software.
Whichever vendor you choose, a BAA delegates obligations to it, but your own liability stays with you.
Which one fits which team
Choose on whether you want to pick the models yourself, then test your shortlist on a real phone line.
Retell suits teams that want to choose and swap language models and need a BAA without signing an Enterprise contract, with up to 20 calls at once included. Bland suits teams that would rather pay one rate that covers the models and are buying Enterprise anyway, with EU data residency or a private deployment written into that contract.
Picking the platform is a small part of the work, though. Most of the effort goes into connecting the agent to your CRM and calendar, and into designing the call, and that work is much the same on either vendor. Our guide to integrating voice agents with your stack covers that side.
Before you sign either contract, run the same call script on both over a real phone line. Most phone calls still reach the agent as 8 kHz audio, while a browser demo runs at 16 kHz, so an agent that sounds sharp in the browser can mishear callers on the phone. Our checklist for testing voice agents before you go live puts those checks in order.
Join the Dograh Community
Dograh is an OSS alternative to Vapi. Join our Slack community for queries, releases, best practices & community interactions.
The open-source option: Dograh
Both offer on-prem only as an Enterprise contract for closed-source software with no published price. Dograh is open source: anyone can self-host it for free, read the code and change it.
Dograh, an open-source, self-hostable voice agent platform, is licensed under BSD-2-Clause, and its code is on GitHub. Dograh is the orchestration layer: it runs the call flow, tools and telephony, and the speech and language models are a separate layer you choose. That layer can be Dograh's own models, your own keys for any supported provider, or open-weight models you host.
| Product | Open source | Self-host or on-prem | Platform fee | What you get, and what's missing |
|---|---|---|---|---|
| Dograh | Yes, BSD-2-Clause | Anyone can self-host it for free; on-prem or private cloud support on Enterprise | Self-hosted: none. Dograh Cloud: 1¢/min | A visual workflow builder, with Dograh's models, your own keys or open-weight models you host. But you run the servers, call transfer to a person works on Twilio only, and no BAA is published |
| Retell AI | No | On-prem through Enterprise sales, no published price or install details | $0.055/min voice infrastructure | Model choice across OpenAI, Anthropic and Google, a visual Conversation Flow builder and a no-fee BAA. But Retell bills the model itself, and the code stays closed |
| Bland AI | No | VPC, on-prem or air-gapped on Enterprise, no published price | None on Start; $299 a month on Build | One rate covering Bland's own models, with the Pathways and Norm builders. But there is no model choice on Start or Build, and the BAA is Enterprise only |
Self-hosted Dograh has no per-minute platform fee, but you still pay for servers and any paid model or phone accounts. Call data stays on your servers only when you self-host Dograh and host the models too. With API keys, the audio reaches the model provider, and the phone carrier carries the call either way.
Self-hosted, you control the server and choose when to upgrade. Dograh Enterprise adds support with service-level agreements (SLAs) if you need it. Our post on the open-source alternative to Retell AI covers that trade in depth, and Dograh vs Retell AI compares the two row by row.
If you are choosing between Retell and Bland, price your own minutes on both pages and test each over a real phone line. If you need the platform on your own servers without an Enterprise contract, run that same test on a self-hosted copy of Dograh.
Glossary
- Business associate agreement (BAA)
- The contract HIPAA requires before a vendor handles patient data. It hands the vendor obligations, but your own liability stays with you.
- Conversational Pathways
- Bland's visual builder. Each node handles one part of the call, and the lines between nodes decide where the call goes next.
- Data residency
- Keeping call data stored and processed inside one chosen region, such as the EU. On Bland it is an Enterprise feature.
- Air-gapped setup
- Running the platform on servers with no connection to the internet, so no call data can leave over the network.
