Why AMD cost is hard to compare
Answering machine detection vendors price in different units. One charges per answered call, another per invocation, another a flat monthly plan, and Asterisk's built-in AMD() charges nothing at all. On top of that, two of the best-known options only work if you place your calls through their platform, which changes what you are actually buying.
To compare fairly you need one unit. This post uses answered calls per month, because that is when detection runs. It also separates the detection fee from call minutes, because minutes are a separate cost in every case.
Twilio and Telnyx prices below are from their public pricing and release notes as of November 2026. Vendors change prices, and high-volume accounts often have negotiated rates. Check each vendor's current pricing page before you decide.
The comparison at 1M, 3M and 10M answered calls
Monthly AMD fee only. Call minutes are excluded in every row.
| Option | 1,000,000 calls | 3,000,000 calls | 10,000,000 calls |
|---|---|---|---|
| Twilio AMD ($0.0075 per answered call) | $7,500 | $22,500 | $75,000 |
| Telnyx premium AMD ($0.0065 per invocation) | $6,500 | $19,500 | $65,000 |
| Telnyx standard AMD ($0.0020 per invocation) | $2,000 | $6,000 | $20,000 |
| Asterisk AMD() (built in) | $0 + hidden costs | $0 + hidden costs | $0 + hidden costs |
| nomoVM | $99 (Starter) | $199 (Growth) | $499 (Scale) |
At 10 million answered calls, the spread between the most and least expensive paid option is $74,501 a month. Even against Telnyx's standard tier, nomoVM is $19,501 a month less at that volume. Remember that Twilio and Telnyx fees come on top of their call minutes, and only apply to calls placed through their platforms.
Cost per 1,000 calls
Per-1,000 pricing makes it easier to compare against your own volumes.
| Option | Cost per 1,000 answered calls |
|---|---|
| Twilio AMD | $7.50 |
| Telnyx premium AMD | $6.50 |
| Telnyx standard AMD | $2.00 |
| nomoVM Starter ($99 for 1,000,000), at full use | about $0.10 |
| nomoVM Growth ($199 for 3,000,000), at full use | about $0.07 |
| nomoVM Scale ($499 for 10,000,000), at full use | about $0.05 |
| nomoVM extra calls on paid plans ($0.0002 each) | $0.20 |
The nomoVM figures assume you use the plan's full allowance. If you use half of it, the effective rate per 1,000 doubles, which is why picking the right plan matters (more on that below). Even nomoVM's overage rate of $0.20 per 1,000 is a tenth of Telnyx's standard tier.
A worked example
Take a hypothetical VICIdial center that answers 2,000,000 calls a month over its own SIP trunks. Here is its answering machine detection cost under each option, before minutes:
- Twilio: 2,000,000 × $0.0075 = $15,000 a month, but only after moving all dialing onto Twilio's API and paying Twilio's call minutes.
- Telnyx premium: 2,000,000 × $0.0065 = $13,000. Telnyx standard: 2,000,000 × $0.0020 = $4,000. Both require rebuilding call control on Telnyx's API; a plain Telnyx SIP trunk doesn't qualify.
- Asterisk AMD(): $0 in fees, plus the tuning time and lost conversations described below.
- nomoVM: Starter would be $99 + (1,000,000 × $0.0002) = $299, so Growth at $199 is the better plan. Trunks, dialer and agents stay as they are.
The fee gap is large, but the bigger difference for this center is that the first two options would mean replacing how it places calls.
What each price actually includes
Price per call is only useful if the products do the same job. They don't, quite.
Twilio AMD
A MachineDetection parameter on calls created through Twilio's API, with the result in AnsweredBy (human, several machine_* values, fax or unknown). It can run asynchronously. It only works for calls placed through Twilio, and your application handles the routing. Twilio's marketing claims 95%+ accuracy. Details in our Twilio AMD pricing post.
Telnyx AMD
The answering_machine_detection option when dialing through Telnyx Call Control, with results delivered as webhook events, in standard and premium tiers. It only applies to calls controlled through Telnyx's API, not to calls your own Asterisk places over a Telnyx SIP trunk. More in Telnyx AMD pricing.
Asterisk AMD()
Free and already installed, and VICIdial uses it by default through extension 8369. It decides with timing rules: initial silence, greeting length, silence after the greeting, word count and word length. Results are HUMAN, MACHINE, NOTSURE or HANGUP. The hidden costs:
- Tuning time. Settings like
greeting(1500 ms by default) andmaximum_number_of_wordstrade live people against machines. Every change needs testing on real calls. - Lost conversations. A person who answers with a long greeting, or waits silently, can be marked as a machine (causes like
LONGGREETINGorINITIALSILENCE). Each one is a lead you paid to reach and then hung up on. See what AMD false positives cost. - Dead air. The shipped
total_analysis_timeis 5000 ms. Time spent listening is silence for the person on the line, which drives hang-ups. Our dead air guide covers this. - No call screening category. Timing rules have no way to label Google Call Screen or iPhone screening as their own case.
nomoVM
Hosted AI AMD that runs on the dialer you already have: VICIdial, GoAutoDial, Asterisk 13+ or FreeSWITCH, over any SIP carrier. The dialer streams the first seconds of each answered call over an encrypted WebSocket and gets one verdict back, usually about 2 seconds after answer. The price includes:
- Verdicts as
AMDSTATUS(HUMAN,MACHINE,FALLBACK) with a cause inAMDCAUSE, includingSCREENING,SILENCE,TONEandBUSY. - Safe fallback: if the service can't be used, Asterisk's AMD() runs for that call and it is never billed.
- A dashboard with call logs, playback, CSV export and usage.
- Recordings of only the seconds the service hears, which you can turn off, keep for 1 to 90 days, or delete at any time.
- No per-server or per-agent fees, and unlimited servers.
We publish both error numbers from a test on 400 real outbound calls checked by a person: 94.9% of live people were sent to agents, and 88% of calls sent to agents were live people. See the accuracy page.
How to calculate your own AMD cost
Count answered calls, not dials
AMD only runs once a call is answered. Take answered calls per month from your dialer reports. If you only know dials, multiply by your answer rate.
Price the per-call vendors
Answered calls × per-call fee. For example, 1,400,000 answered calls × $0.0075 = $10,500 on Twilio, or × $0.0020 = $2,800 on Telnyx standard. Then add their call minutes, since you would be placing calls through them.
Price nomoVM
Plan price + (calls over the included amount × $0.0002). At 1,400,000 calls, Starter is $99 + (400,000 × $0.0002) = $179, which beats Growth at $199.
Find the plan break-even points
Starter plus overage equals Growth at 1,500,000 calls ($99 + $100 = $199). Growth plus overage equals Scale at 4,500,000 calls ($199 + $300 = $499). Below those points, the smaller plan is cheaper; above them, move up. Above 10 million a month, talk to us about Enterprise.
Add the cost of wrong answers
Fees are only part of it. Estimate how many live people your current AMD sends away and what a conversation is worth. The ROI calculator does this in your browser.
Which option fits your setup
- Your app already places calls through Twilio or Telnyx's API: their built-in AMD is the shortest path. Price it at your real volume with the tables above.
- You run VICIdial, GoAutoDial, Asterisk or FreeSWITCH on your own trunks: Twilio and Telnyx AMD aren't available on those calls. Your choice is between Asterisk AMD() and a service that runs on your dialer.
- You want better results than AMD() without new infrastructure: nomoVM adds one script on the dialer you already run. No new servers, no hardware, no carrier change, and about 10 minutes to set up. Compare it in detail on our comparison page.
To see the numbers on your own calls, open a free account (30,000 calls a month, no card), follow the VICIdial AMD setup guide, and run one campaign for a week. Plan details are on the pricing page.
Frequently asked questions
What is the cheapest answering machine detection?
On fees alone, Asterisk's built-in AMD() is free, but it costs you in tuning time, lost live conversations and dead air. Among paid options, nomoVM's flat plans cost less per call than Twilio or Telnyx at every volume in our comparison, for example $99 a month for 1,000,000 calls.
How much does answering machine detection cost per 1,000 calls?
Using public prices from November 2026: Twilio $7.50, Telnyx premium $6.50 and Telnyx standard $2.00 per 1,000, all on top of call minutes. nomoVM works out to about $0.10 per 1,000 on Starter at full use, falling to about $0.05 on Scale.
Are AMD fees included in call minutes?
No. Twilio and Telnyx charge their AMD fees on top of call charges, and only on calls placed through their platforms. nomoVM doesn't carry calls at all, so you keep paying your existing carrier for minutes and pay nomoVM only for detection.
Is Asterisk AMD really free?
There is no license fee, but it is not free to run. Someone has to tune its timing rules, it sends some live people away as machines, and its listening time adds silence before a person reaches an agent. Those costs rarely show on an invoice but they are real.
How do I estimate my monthly AMD cost?
Start with answered calls per month, not dials, because AMD runs on answered calls. Multiply by the per-call fee for per-call vendors. For nomoVM, take the plan price and add $0.0002 for each call over the included amount, then pick the cheapest plan.
Do nomoVM prices change with servers or agents?
No. nomoVM charges per verdict on an answered call, with no per-server or per-agent fees and no limit on servers. Fallback calls, health checks and unanswered calls are never billed.