Millis AI Alternatives (2026): 7 Platforms Tested
Millis AI is one of the cheapest voice agent platforms we've tested: a $0.02-a-minute platform fee plus the LLM, voice and speech-to-text you pick, and about $0.09 a minute billed on our own default test calls. It replied in a median 1.25 seconds on real phone calls and passed 12 of 15 calls in our receptionist test. People usually look elsewhere for what Millis doesn't publish: a limit on calls at once, uptime terms, free trial credits, a HIPAA BAA, or even a pricing page on its main site. Here are seven alternatives, each priced from its own pricing page and each tested by Cozzo on the same phone calls.
Cozzo tip: if your Millis agent runs on your own Twilio, Vonage or Plivo number, that number stays in your carrier account, so you can point it at a new platform and test side by side before you switch.
ElevenLabs Agents, Ultravox, Retell, Vapi, Bland, LiveKit and OpenAI Realtime vs Millis AI: official prices plus Cozzo's own tests on real phone calls.
Which one to start with
ElevenLabs AgentsAgents: $0.08 per minute1.06 s median reply in Round 4, fastest of the 8 platforms, and 1.04 s on our Oct 20 phone calls. About $0.08 a minute in plan credits on our calls.IfYou want one flat price that includes the voice
Ultravox$0.05/min$0.05 a minute with its own model and voices, first 30 minutes free. Passed 14 of 15 Round 4 calls, the most of any platform we tested.IfYou need published limits and a HIPAA BAA
Retell AI$0 platform fee20 calls at once included, $10 in free credits, and a BAA you can self-sign online. Passed 12 of 15 Round 4 calls, the same as Millis.IfYou want to choose every model and carrier
Vapi$0 to start$0.05 a minute hosting with models at cost, so it's the closest to Millis's pick-your-parts pricing. About $0.08–$0.09 a minute on our calls.IfYou want one all-in rate with no model bills
Bland AI$0.14/min, $0 platform fee$0.14 a minute on Start with the LLM, speech-to-text and voice included. Slowest stop after a cut-in in Round 4 (2.68 s), so test interruptions.IfYou'd rather own the stack
LiveKit AgentsFreeOpen-source framework you can self-host, or LiveKit Cloud with 1,000 free agent minutes a month. Our self-hosted build passed 11 of 15 Round 4 calls.Millis AI vs each alternative, at a glance
- MMillis AIYou're here
- Published price
- $0.02/min platform + LLM, voice and speech-to-text (docs.millis.ai)
- Phone reply
- 1.25 s12 calls, Oct 20
- Round 4 passed
- 12 of 151.31 s median
- Cost on our calls
- about $0.09/min
- Why pick it
- Low bill if you pick a cheap voice; limits and compliance not published
- ElevenLabs Agents
- Published price
- $0.08/min, LLM billed on top; plans include 15 to 12,375 minutes
- Phone reply
- 1.04 s11 calls, Oct 20
- Round 4 passed
- 13 of 151.06 s median
- Cost on our calls
- about $0.08/min equivalent
- Why pick it
- Fastest replies in Round 4, and the company behind the ElevenLabs voices Millis offers
- Ultravox
- Published price
- $0.05/min, first 30 minutes free; model and voices included
- Phone reply
- 1.18 s11 calls, Oct 20
- Round 4 passed
- 14 of 151.27 s median
- Cost on our calls
- $0.05/min list price
- Why pick it
- Most Round 4 calls passed (14 of 15) at one flat rate
- Retell AI
- Published price
- $0.07–$0.31/min; $10 free credits; 20 calls at once included
- Phone reply
- 1.04 s11 calls, Oct 20
- Round 4 passed
- 12 of 151.38 s median
- Cost on our calls
- about $0.15/min
- Why pick it
- Published limits and a BAA you can self-sign
- Vapi
- Published price
- $0.05/min hosting, models at cost; $5 free credits; HIPAA add-on $2,000/mo
- Phone reply
- 1.56 s11 calls, Oct 20
- Round 4 passed
- 9 of 151.90 s‡ median
- Cost on our calls
- about $0.08–0.09/min
- Why pick it
- Pick every model and carrier yourself
- Bland AI
- Published price
- $0.14/min on Start; $0.12/min on Build ($299/mo); LLM and voice included
- Phone reply
- 1.86 s11 calls, Oct 20
- Round 4 passed
- 10 of 151.94 s median
- Cost on our calls
- about $0.14/min
- Why pick it
- One all-in rate, no separate model bills
- LiveKit Agents
- Published price
- Open source; LiveKit Cloud free for 1,000 agent minutes, Ship $50/mo
- Phone reply
- 1.52 s11 calls, Oct 20
- Round 4 passed
- 11 of 152.00 s median
- Cost on our calls
- Not measured
- Why pick it
- Own the whole stack or self-host
OpenAI Realtime API- Published price
- gpt-realtime-2.1 audio $32 in / $64 out per 1M tokens; mini $10 / $20
- Phone reply
- 1.52 s11 calls, Oct 20
- Round 4 passed
- 12 of 151.71 s median
- Cost on our calls
- Not measured
- Why pick it
- Build straight on OpenAI's speech-to-speech model
Prices from each vendor's own pricing page, read October 25, 2026. Measured columns are Cozzo's own calls on each vendor's default model and voice: phone reply = median on our Oct 20 phone calls; Round 4 = calls passed out of 15 on our Oct 22 receptionist test; cost = what each dashboard billed us. Small samples, so treat close numbers as ties. ‡ Vapi's Round 4 median is end-to-end on an untuned default setup, not comparable to Vapi's published figures. See the phone-call benchmark and the AI receptionist benchmark for methods.
- ElevenLabs Agents
- Starting price
- Agents: $0.08 per minute
- Best for
- Builders who want ElevenLabs' voices and speech models in one agent pipeline
- Ultravox
- Starting price
- $0.05/min (first 30 min free)
- Best for
- Developers wanting a speech-native model at a flat per-minute rate
- Retell AI
- Starting price
- $0 platform fee; $0.07–$0.31/min for voice agents
- Best for
- Teams that want a no-code flow builder plus an API with itemized per-minute pricing
- Vapi
- Starting price
- $0 to start; $0.05/min Vapi hosting, models billed pass-through at cost
- Best for
- Developers who want to choose their own LLM, transcriber, voice and telephony provider
- Bland AI
- Starting price
- $0.14/min, $0 platform fee
- Best for
- Teams that want one flat per-minute price with no model pass-through bills
- LiveKit Agents
- Starting price
- Free; Ship from $50/mo
- Best for
- Developers building on open-source realtime infrastructure
- OpenAI Realtime API
- Starting price
- gpt-realtime-2.1-mini: audio $10 in / $20 out per 1M tokens
- Best for
- Developers building speech-to-speech voice agents directly on OpenAI models
Pricing explained, line by line:ElevenLabs Agents pricing →Ultravox pricing →Retell AI pricing →Vapi pricing →Bland AI pricing →
Cozzo-tested · Round 4 · October 22, 2026
AI receptionist benchmark: 8 voice agents on 5 real-call scenarios
Millis AI passed 12 of 15 calls here. Every alternative on this page ran the same dental-office script, 3 calls per scenario. Each chip shows calls passed out of 3. Fastest median reply first.
ElevenLabs · 13/15 · 1.06 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
Ultravox · 14/15 · 1.27 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
Millis AI · 12/15 · 1.31 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
Retell AI · 12/15 · 1.38 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
OpenAI Realtime · 12/15 · 1.71 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
Vapi · 9/15 · 1.90 s‡ median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
Bland AI · 10/15 · 1.94 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
LiveKit · 11/15 · 2.00 s median reply
- Book a new patient
- Reschedule
- Unknown insurance
- Interruption
- Dental emergency
‡ Vapi’s 1.90 s median is end-to-end on an untuned default setup, not comparable to Vapi’s published figures, which leave out endpointing and network time. † Callback taken, but the number was read back wrong at least once. * Reply heard each time, but its content wasn’t checked against the vendor’s own transcript. 3 calls per scenario, one prompt, English only: small samples, so treat close numbers as ties.
Cozzo tested: reply time on real phone calls
Median time from the end of the caller’s speech to the agent’s first reply, on each vendor’s default model and voice (Oct 20, 2026). Places are out of 8 platforms tested.
- ElevenLabs Agents: 1.04 s (tied 1 of 8, 11 calls)
- Retell AI: 1.04 s (tied 1 of 8, 11 calls)
- Ultravox: 1.18 s (3 of 8, 11 calls)
- LiveKit Agents: 1.52 s (tied 5 of 8, 11 calls)
- OpenAI Realtime API: 1.52 s (tied 5 of 8, 11 calls)
- Vapi: 1.56 s (7 of 8, 11 calls)
- Bland AI: 1.86 s (8 of 8, 11 calls)
See the full phone-call benchmark → Small samples, so treat close numbers as ties.
What to check before you choose
- Price the whole minute, not the headline. Millis's $0.02 is only the platform fee: with an ElevenLabs voice its own docs put a minute at about $0.078, and our default calls billed about $0.09. Compare that with Ultravox's flat $0.05, Vapi's $0.05 hosting plus models, Retell's $0.07 to $0.31, ElevenLabs Agents' $0.08 plus the LLM, or Bland's $0.14.
- Keep your phone number. Millis works with a number you import from Twilio, Vonage, Plivo or Exotel, or with SIP. Because that number lives in your carrier account, Retell, Vapi, ElevenLabs Agents, Ultravox and LiveKit can all take it over without porting.
- Rebuild the prompt and tools first, then compare. Copy your Millis system prompt, webhooks and function calls into the new platform and run the same test script on both before you move live traffic.
- Test the callback read-back. Round 4's booking test needed the caller's number read back exactly, and no platform passed all 3 calls; Millis passed 0 of 3. If your callers leave numbers, make your agent read them back in groups and check the transcripts.
- Ask about paperwork before patient or card data. Retell lets you self-sign a BAA; Vapi sells HIPAA as a $2,000-a-month add-on; ElevenLabs, Ultravox and Bland offer BAAs on enterprise terms; LiveKit lists a signed BAA from its Scale plan ($500 a month).
- Check concurrency. Retell includes 20 calls at once, Vapi 4 on its free tier, Bland 10 on Start and LiveKit Cloud 5 on its free Build plan. Millis doesn't publish a limit.
Questions people ask
What are the best Millis AI alternatives?
On Cozzo's own calls the strongest were ElevenLabs Agents (fastest replies: 1.06 s median in Round 4), Ultravox (most receptionist calls passed, 14 of 15, at a flat $0.05 a minute) and Retell AI (12 of 15, with published limits and a self-sign BAA). Vapi suits builders who want to pick every model, Bland AI suits teams that want one all-in rate, and LiveKit Agents or the OpenAI Realtime API suit developers who want to build their own stack.
Is there a cheaper alternative to Millis AI?
Ultravox is the simplest cheaper option on list price: $0.05 a minute with its model and voices included, against the roughly $0.09 a minute Millis billed on our default calls (its default LLM with an ElevenLabs voice). Millis itself can be cheaper than that if you pick an OpenAI or Deepgram voice: its own docs put that mix at about $0.036 a minute. Phone carrier charges are extra on all of them.
Which Millis AI alternative is fastest?
ElevenLabs Agents. It had the fastest median reply in our Round 4 receptionist calls (1.06 s vs Millis's 1.31 s) and tied Retell AI for fastest on our Oct 20 phone calls (1.04 s vs Millis's 1.25 s). These are end-to-end times on real phone lines with each vendor's default setup.
Which Millis AI alternative is HIPAA-ready?
Millis doesn't publish HIPAA, SOC 2 or BAA terms. Retell AI says it's HIPAA compliant and lets you self-sign a BAA. Vapi sells HIPAA as a $2,000-a-month add-on. ElevenLabs, Ultravox and Bland AI offer BAAs on enterprise terms, and LiveKit lists a signed BAA on its Scale plan ($500 a month) and up.
Can I keep my phone number if I leave Millis AI?
Usually, yes. Millis uses numbers you import from Twilio, Vonage, Plivo or Exotel, or calls routed over SIP, so the number stays in your carrier account. Point it at the new platform's SIP address or import it there. If Millis bought the number for you, ask Millis how to move it.
How did Cozzo test these platforms?
We called each platform on its default model and voice over real phone lines. On Oct 20 we measured reply time and transcription on about a dozen calls each; on Oct 22 (Round 4) we ran 5 dental-receptionist scenarios, 3 calls each; and on Oct 23 (Round 5) we ran gym, coaching and roofing scripts. Costs are what each vendor's dashboard billed us. Full methods are on the benchmark pages.
Is Millis AI still worth using?
Yes, if a low per-minute bill matters most and you can live without published limits. Millis passed 12 of 15 Round 4 calls, the same as Retell AI, stopped within 0.60 s when a caller cut in, and did well on Round 5's gym and coaching calls. Ask Millis about concurrency, uptime and a BAA before you scale.
Already leaning toward Millis AI? See how Millis AI’s pricing works, line by line →
Not sure which fits? Answer a few quick questions in the finder →
Prices from vendors' official pricing pages, checked October 25, 2026. Not sponsored; order isn't a ranking.