Telnyx voice AI pricing explained: what the $0.05 rate covers, and what bills on top

Telnyx advertises its voice-AI agents at $0.05 per minute, with the headline “$0.05 per minute. No fine print.” The rate combines orchestration, speech-to-text, and text-to-speech. However, the component table on the same page separately itemizes language-model tokens and telephony, so a configured call can cost more than the headline rate. Telnyx voice AI pricing, accessed 2026-09-01.

This distinction is the key to understanding Telnyx voice AI pricing. The 5-cent figure is a real published bundle, but it is not an LLM-and-telephony-inclusive total. Telnyx’s own realistic example lands at approximately $0.056 per minute, combining the voice engine with estimated LLM usage and inbound US local telephony. This guide follows Telnyx’s own itemization without treating either figure as a universal quote.

How the Telnyx voice AI pricing model works

The $0.05/minute voice-engine rate covers the conversational layer: agent runtime and orchestration, including turn-taking, interruptions, tool use, and knowledge retrieval, plus speech-to-text and text-to-speech. The pricing page presents its supported STT and TTS options within that bundled rate rather than assigning a separate per-minute charge to each option. Telnyx voice AI pricing, accessed 2026-09-01.

Two major call components then bill separately:

• Language-model usage. Models running on Telnyx-owned GPUs are described as “typically ~$0.004/min.” Anthropic and OpenAI models instead follow their respective token rates. Actual billing is per token, and Telnyx uses roughly 6,000 prompt tokens per minute when converting token prices into estimated per-minute figures.
• Telephony. Inbound US local calling starts at $0.0032/minute, while outbound SIP starts at $0.005/minute. These are starting rates, so destination and call direction matter.

Both sets of figures come from Telnyx voice AI pricing, accessed 2026-09-01.

Telnyx illustrates the structure with a realistic configuration totaling about $0.056/minute and approximately $12,968/month at 230,000 minutes. Its example uses the $0.05 voice engine, $0.0032 inbound telephony, and roughly $0.0032 of LLM usage. Those are Telnyx’s published approximations, not a customer-specific quote.

There is a small rounding detail worth making explicit. Adding the displayed component estimates gives $0.0500 + $0.0032 + $0.0032 = $0.0564, which Telnyx presents as approximately $0.056. At 230,000 minutes, the displayed components compute to 230,000 × $0.0564 = $12,972, close to the page’s approximately $12,968 monthly figure. Using the rounded $0.056 alone would instead produce 230,000 × $0.056 = $12,880. These computations are illustrative arithmetic, not quotes; token billing uses actual tokens rather than the rounded planning estimate.

What is included and what can bill separately

The practical lesson is to estimate a complete call rather than multiplying the headline rate alone. Model choice changes the token line, direction changes telephony, and optional recording or Voice API charges can add further usage costs.

Worked example: 10,000 minutes a month

Consider 2,000 calls × 5 connected minutes = 10,000 connected minutes per month. Assume inbound US local calls and Telnyx’s own realistic rounded estimate of $0.056/minute. Every computation below is illustrative planning arithmetic, not a Telnyx quote.

Voice agent, estimated LLM, and inbound telephony:
10,000 × $0.056 = $560/month

Using the displayed components before rounding produces a slightly different planning result:

10,000 × ($0.0500 + $0.0032 + $0.0032)
= 10,000 × $0.0564
= $564/month

If every minute is recorded and one US local number is used:

Recording: 10,000 × $0.002 = $20/month
Number: $1/month
Total using Telnyx’s rounded $0.056 estimate: $560 + $20 + $1 = $581/month

If the separately published $0.002/minute Voice API platform fee also applies to the chosen call path, that adds another 10,000 × $0.002 = $20/month, producing $581 + $20 = $601/month. Confirm that line’s applicability when configuring the deployment. The underlying rates are from Telnyx voice AI pricing, accessed 2026-09-01.

Actual LLM billing is token-based, so $560, $564, $581, and $601 are planning scenarios rather than invoice predictions. A model with a different token rate, longer prompts, outbound calling, recording choices, or a different destination can move the result.

Who Telnyx voice AI fits

Telnyx positions the product as a developer- and agent-native platform. Its APIs are central, while a no-code AI Assistant Builder is also available in the portal. That combination can suit teams that want an integrated carrier and voice-agent environment while retaining control over models and call infrastructure.

Its published carrier-side starting rates are notable: inbound US local telephony begins at $0.0032/minute and outbound SIP at $0.005/minute. Telnyx voice AI pricing, accessed 2026-09-01. The tradeoff is that buyers need to reason about multiple pricing dimensions. The LLM is not hidden inside the 5-cent engine rate, and selecting a different model changes the token-billed portion of the call.

Gotcha checklist

• “All in” has a specific scope here. The $0.05 rate combines orchestration, STT, and TTS. LLM tokens and telephony remain separate according to Telnyx’s component table.
• Per-minute LLM figures are estimates. Telnyx bills actual tokens and uses roughly 6,000 prompt tokens per minute for its conversion. Prompt size and model selection can move the result.
• Telephony prices are starting rates. The $0.0032/minute figure is for inbound US local calling, while outbound SIP starts at $0.005/minute. Do not assume every destination shares those rates.
• Small component lines still matter at volume. The published Voice API platform fee and recording charge are each $0.002/minute where applicable. At 10,000 minutes, each illustratively computes to 10,000 × $0.002 = $20, not a quote.
• Volume discounts require commitment. Pay as you go has no commitment, while the Committed and Enterprise tiers begin at $500 and $5,000 monthly minimums, respectively.
• Pretrial credit changes on upgrade. The $25 pretrial credit is for AI products; upgrading replaces it with a $5 general credit rather than adding the two balances together.

All checklist figures are from Telnyx voice AI pricing and Telnyx pretrial accounts, accessed 2026-09-01.

How ThunderPhone pricing differs

ThunderPhone uses one complete engine rate that includes speech recognition, the language model, and voice: Spark at 2¢/minute, Bolt at 5¢/minute, and Storm at 9¢/minute. Pricing is prepaid and pay as you go, with no subscription, seat fee, or platform fee. Telephony depends on setup.

For the same 10,000-minute workload, the illustrative arithmetic, not a quote, is:

• Spark: 10,000 × $0.02 = $200/month
• Bolt: 10,000 × $0.05 = $500/month
• Storm: 10,000 × $0.09 = $900/month
• Telnyx’s rounded realistic estimate: 10,000 × $0.056 = approximately $560/month

This is not a universal ranking. Telnyx’s estimate includes its starting inbound telephony figure, while ThunderPhone telephony depends on setup; Telnyx and ThunderPhone also target different kinds of builders. Call length, language, model choice, telephony, and configuration can reorder results, so rerun the arithmetic for the actual deployment.

For broader context, see the cheapest AI voice-agent guide, the complete ThunderPhone pricing guide, and the cross-platform pricing comparison.

FAQ

Is Telnyx really $0.05 per minute all in?

The $0.05 rate includes orchestration, speech-to-text, and text-to-speech; LLM tokens and telephony bill separately. Telnyx’s own realistic example totals approximately $0.056/minute after estimated LLM usage and inbound US local telephony. Telnyx voice AI pricing, accessed 2026-09-01.

What does a Telnyx voice agent cost per month?

It depends on minutes, model tokens, telephony, and optional services. At Telnyx’s rounded realistic estimate, 10,000 minutes illustratively compute to 10,000 × $0.056 = $560/month, not a quote. Recording every minute and adding one US local number would make that $560 + $20 + $1 = $581/month; an applicable Voice API platform fee could add another $20.

Does Telnyx have a free tier?

Telnyx publishes pretrial credit rather than a permanent free voice-AI tier in the cited materials. The pretrial includes $25 in AI credits, requires no card, and includes one US local number. On upgrade, that balance is replaced by a $5 credit usable across products. Telnyx pretrial accounts, accessed 2026-09-01.

How does Telnyx compare with Retell, Vapi, and ThunderPhone on price?

Compare the complete configured minute, because the vendors’ headline rates cover different bundles. Telnyx’s $0.05 includes orchestration, STT, and TTS but excludes LLM tokens and telephony; ThunderPhone’s engine rates include STT, the LLM, and voice, with telephony depending on setup. Retell and Vapi use their own component or pass-through structures. See the normalized voice AI pricing comparison, then rerun it for your call length, language, and configuration.

Sources and freshness

All Telnyx claims and figures were checked against Telnyx voice AI agent pricing and Telnyx’s pretrial-account documentation, accessed 2026-09-01. ThunderPhone pricing reflects its published product information as of 2026-09-01. This guide is documentation-based; no Telnyx invoice or live call was tested. Vendor rates, token assumptions, destinations, and tier terms can change, so confirm the live pages and reproduce the arithmetic before making a purchasing decision.