Voice agent platforms at 10,000 minutes: eight price lists compared
Telnyx about $596, Deepgram and AssemblyAI $750, Bland $1,400 to $1,499, Retell $700 to $3,100: eight published rates at 10,000 minutes.
At 10,000 minutes a month the published voice agent prices run from about $596 on Telnyx to $3,100 at the top of Retell's range, and the ranking depends on which lines each vendor includes. Telnyx, Deepgram's standard tier and AssemblyAI bundle the speech steps; Vapi, Retell and LiveKit price them as separate lines. Sume sells none of these: it has no live agent. Its audio steps are jobs, so it appears here only as the cost of the recorded pieces.
Every row is from the vendor's pricing page, read 2026-10-03: Vapi, Retell, Bland, Telnyx, Deepgram, AssemblyAI, LiveKit and ElevenLabs. Monthly figures are my multiplication.
The table
The rows are not like for like. Read the notes column before comparing.
| Platform | Listed rate per minute | 10,000 minutes | Note |
|---|---|---|---|
| Telnyx Voice AI | about $0.06 | about $596 | Engine $0.05 plus LLM and inbound SIP |
| Deepgram Voice Agent, standard | $0.075 | $750 | Advanced tier is $0.163, $1,630 |
| AssemblyAI Voice Agent API | $4.50 an hour | $750 | Described as all-inclusive |
| Bland Start | $0.14 | $1,400 | Caps of 100 calls a day |
| Bland Build | $0.12 + $299 a month | $1,499 | 50 concurrent calls |
| Vapi, hosting only | $0.05 | $500 | STT, model, voice and transport extra |
| Retell, standard range | $0.07 to $0.31 | $700 to $3,100 | Built from separate lines |
| ElevenLabs Speech Engine | $0.08 beyond included | $800 if none included | Burst price is $0.16 |
Reading it fairly
Vapi's $500 is hosting only; its page lists a transcriber at $0.0095 to $0.0099, a model at $0.0077 to $0.0452 and a voice at $0.0146 to $0.0238 a minute, which brings the total to about $818 to $1,289 for 10,000 minutes by my addition. LiveKit lists agent sessions at $0.01 a minute with STT, TTS and LLM ranges on top and is omitted from the table for that reason.
The ElevenLabs row is the least certain: its page shows $0.080 for each additional minute beyond an included allotment, and I did not see the allotment. Treat $800 as an upper bound for this volume only if nothing is included.
One thousand and one hundred thousand minutes
The ranking at 10,000 minutes does not hold at the extremes, because Bland has a platform fee and Vapi's Pro package has a $999 minimum.
| Platform | 1,000 minutes | 100,000 minutes |
|---|---|---|
| Telnyx, about $0.06 | $60 | $5,960 |
| Deepgram standard, $0.075 | $75 | $7,500 |
| Bland Start, $0.14 | $140 | $14,000 |
| Bland Build, $0.12 + $299 | $419 | $12,299 |
What Sume does and does not do
The recorded pieces are cheap beside a live minute. A 150-character greeting rendered once is one cent at $0.0475 per 1,000 characters, rounded up. A call recording transcribed is $0.01 a minute. Neither replaces the live meter; both reduce what you route through it. See pre-rendering fixed replies and the voice agent stack.
Sources
Related posts
More in Comparisons
- Voice isolator vs audio detach: 500 MB and 1 hour vs 1800 seconds
ElevenLabs Voice Isolator removes background noise from speech. Sume audio detach extracts the video audio track as is; it does not separate vocals from music.
- Wan 3.0 or MiniMax H3 Max: which to pin for reference-to-video
Both take image, video and audio references on Sume. Wan runs 2 to 30 seconds with 5 reference videos; H3 Max runs 5 to 15 seconds with always-on stereo audio.
- Wan 3.0 or Seedance 2.5: which is cheaper for a 10-second 720p clip?
Wan 3.0 is cheaper: a 10-second 720p clip is $1.00 at fal list ($1.25 on Sume) against about $4.62 ($5.78 on Sume) for Seedance 2.5, roughly 4.6x.
- Wan 3.0 vs Gemini Omni Flash at 720p: both list $0.10 per second
Wan 3.0 (fal) and Gemini Omni Flash (Google, about $0.10 per second at 720p) share a list price; Sume bills $0.125 per second for either.
Written by Sume