Voice agents priced at $0.08 per call minute with no telephony fee from ElevenLabs, but LLM tokens billed separately at cost on top.
Teams where voice quality and response speed are the deciding factors, since ElevenLabs publishes model-level latency figures no orchestration vendor matches: about 75ms for Flash v2.5 and about 100ms median inference for Eleven v4 Turbo, excluding application and network latency. The published tier table makes capacity planning straightforward, with concurrency stated per plan from 4 calls on Free to 40 on the $990 Business plan. It suits SIP-first deployments too, as the docs name Twilio, Vonage, Telnyx, RingCentral, Sinch, Infobip, Exotel, Plivo and Bandwidth.
Buyers expecting an all-in rate should note the FAQ states LLM fees are billed separately, vary by model and are charged at cost, so $0.08 a minute is a floor rather than a total. Concurrency is hard-tiered and overflow is expensive: burst allows up to three times the plan limit but charges double, at $0.16 a minute. Multilingual coverage on the agents platform is 31 languages and the docs state language selection is fixed for the duration of a call, so a caller cannot switch mid-conversation. The free plan is 15 call minutes, enough to test and nothing more.
Pros
Cons
It covers the agent call minute itself, including ElevenLabs speech synthesis and transcription, and ElevenLabs adds no telephony fee. LLM usage is not included: the FAQ states LLM fees are billed separately, vary by model and are charged at cost. Carrier charges are also separate, billed directly by Twilio or whichever SIP provider you connect to the agent.
Burst pricing applies. Agents can handle up to three times the plan concurrency limit, and every call above the limit is charged at double the standard rate, which is $0.16 per minute instead of $0.08. For inbound SIP traffic the docs state callers are rejected once the limit is reached unless call queueing is enabled on the agent, in which case they wait for capacity.
Yes, at the model level. The models page lists about 75ms for Flash v2.5, about 100ms median inference for Eleven v4 Turbo, about 150ms for Scribe v2 Realtime speech-to-text and about 280ms for Eleven v3 Conversational. ElevenLabs states these exclude application and network latency, so they are component figures rather than an end-to-end conversational round trip.
No. The agents platform supports 31 languages and a language detection tool can route a caller to their preferred language, but the docs state language selection is fixed for the duration of the call. English agents default to the Flash v2 model, and adding any additional language switches the agent to the v2.5 Multilingual model instead.
Facts sourced from the vendor’s own site, last verified 2026-10-10.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
