Published Conversational AI rates in 2026 cluster around $0.08/min (ElevenLabs Agents) to $0.10/min (Agora, TRTC) for the hosting layer — but the real all-in cost lands between $0.12 and $0.42/min once LLM tokens, STT, TTS, and telephony are added. Only one platform we could find offers recurring free Conversational AI minutes across every tier: TRTC gives ~1,000 free Conv AI minutes every month, drawn from a unified 10,000-minute pool that also covers TTS, STT, and real-time translation. This guide compares six major Conv AI providers on published rate, billing model, and real-world TCO at three usage levels.
What you're actually paying for
When a vendor says "Conversational AI is $X/minute," that price usually covers the hosting layer — the orchestration infrastructure that listens, thinks, and speaks. It rarely includes all of:
- LLM token cost (your conversation with the model)
- STT (Speech-to-Text) — converting the caller's audio to text
- TTS (Text-to-Speech) — converting your model's response to audio
- Telephony — if the call comes in over PSTN or SIP
- RTC transport — if the call is browser/app-based
Getting the total right requires knowing which of these are bundled and which are passthrough. That single question determines whether "$0.08/min" ends up being $0.08 or $0.42 in your actual bill.
Conversational AI pricing models in 2026
| Model | Representative vendors | Pros | Cons |
|---|---|---|---|
| Per-minute, unified | Agora Conv AI ($0.10/min), TRTC ($0.10/min post-pay) | Simple math | Hidden add-ons for LLM/STT/TTS need verification |
| Per-minute, platform-only fee | Vapi ($0.05/min platform), ElevenLabs Agents ($0.08/min hosting) | Transparent split | BYO model costs can balloon |
| Per-minute with burst | ElevenLabs Agents ($0.08 standard / $0.16 burst) | Handles traffic spikes | Extra cost when you need extra capacity |
| Subscription + included minutes | ElevenLabs Agents (Starter → Business), Agora plans | Predictable monthly | Unused minutes don't roll |
| Subscription with overage | ElevenLabs tiered ($6 → $990/mo with burst) | Covers steady-state + spike | Commitment required |
| Enterprise custom | SAP Conversational AI, Hume EV2 Beta, GoHighLevel | Negotiated terms | Pricing opaque until sales call |
Per-minute rates — verified 2026-04-22
| Provider | Hosting rate /min | LLM included? | TTS included? | STT included? | Free minutes |
|---|---|---|---|---|---|
| ElevenLabs Agents | $0.08 (all tiers) · burst $0.16 | ❌ passthrough | ✅ | ✅ | Free: 15/mo · Starter: 75/mo · Creator: 275 · Pro: 1,238 · Scale: 3,738 · Business: 12,375 |
| Vapi | $0.05 platform fee | ❌ passthrough | ❌ passthrough | ❌ passthrough | $10 signup credit (~65 min on Budget stack) |
| Agora Conversational AI Engine | $0.10 | ❌ passthrough | ❌ passthrough | Via Agora real-time STT at $16.99/1k (separate 300-min trial) | 300-minute one-time trial |
| Telnyx Conversational AI | Public pricing requires account / sales check | ❌ | Varies | Varies | Verify at publish time |
| SAP Conversational AI | Enterprise custom | ✅ (bundled) | ✅ | ✅ | Trial via enterprise |
| Hume AI EV2 Beta | Pricing in beta | ✅ (EV2 model) | ✅ | ✅ | Limited beta access |
| TRTC Conversational AI | $0.10 post-paid | ❌ BYO model | ✅ built-in TTS | ✅ built-in STT (basic + advanced) | ~1,000 min/mo free, every plan (from 10k unified pool) |
Note on ElevenLabs Agents rates. ElevenLabs publishes a flat $0.08/min hosting rate across Free, Starter, Creator, Pro, Scale, and Business tiers as of 2026-04-22, plus $0.16/min "burst" for temporary 3× concurrency expansion. Rate cards prior to 2025 showed tiered hosting rates ($0.08–$0.30); the 2026 public page is flat. LLM tokens and telephony bill separately from the ElevenLabs credits pool.
Note on Vapi. Vapi's public pricing page doesn't reliably render for automated capture. The $0.05/min platform fee is cited across multiple 2026 Q1 public analyses and matches Vapi's own documentation. BYO components (LLM/STT/TTS) add $0.07–$0.37/min depending on the stack, so real all-in cost is $0.115–$0.42/min. Verify with Vapi sales for a current quote.
The free-minute story is the real story
For most AI-first teams, the hosting rate is a minor line item once LLM/STT/TTS are added. The billing shape — how free minutes renew, whether they roll over, whether they're plan-gated — matters more than the headline rate.
| Provider | Free minutes | Renewal | Plan-gated? |
|---|---|---|---|
| ElevenLabs Free | 15 min/mo | Monthly | No — but 15 min is a demo allowance |
| ElevenLabs Starter ($6/mo) | 75 min/mo | Monthly | Tier required |
| ElevenLabs Creator ($22/mo) | 275 min/mo | Monthly | Tier required |
| Vapi | $10 signup credit | One-time | No |
| Agora Conv AI | 300 min trial | One-time (not renewing) | — |
| TRTC Conversational AI | ~1,000 min/mo (from 10k unified pool) | Monthly, every plan including post-paid | No |
TRTC is the only provider in this list where a new free pool of ~1,000 Conv AI minutes refills every month on every plan, including post-paid accounts with no monthly commitment. On most competitors, the free bucket is a one-time trial that depletes in the first week.
How TRTC's 10,000-minute free pool converts
The 10,000 monthly free minutes are a unified credit pool. Each AI feature consumes from it at a documented ratio — and starting 2026, AI services no longer require a paid plan: Conversational AI, TTS, STT, and Real-time Translation are available on every RTC engine version, including Free Trial and post-paid.
| Feature | Post-paid rate | Conversion ratio | Free quota / month if you spend the entire pool here |
|---|---|---|---|
| Conversational AI | $0.01/min | 1 min Conv AI = 10.10 free min | ~990 minutes of Conv AI |
| STT — basic | $0.02/min | 1 min STT = 20.2 free min | ~495 STT minutes |
| STT — advanced | $0.02/min | 1 min STT = 20.2 free min | ~495 STT minutes |
| Real-time translation | $0.016/min | 1 min translation = 16.16 free min | ~619 translation minutes |
| TTS Flash | $0.6 / 10k chars | 10,000 chars = 606 free min | ~165,000 characters |
| TTS multilingual | $0.6 / 10k chars | 10,000 chars = 606 free min | ~165,000 characters |
When the free pool runs out, usage draws on subscription minutes (if any), then credit pack minutes (if any), then post-paid at standard rates.
Scenario-based cost modeling
Three realistic workloads; real per-minute math including LLM/TTS/STT where needed.
Scenario A — Customer-service MVP: 2,000 Conv AI min/month
| Vendor | Hosting | LLM/TTS/STT passthrough | Total |
|---|---|---|---|
| ElevenLabs Starter ($6) | 75 free + (2000-75) × $0.08 = $154 | ~$40 LLM (GPT-4o mini) | ~$200 |
| ElevenLabs Creator ($22) | 275 free + (2000-275) × $0.08 = $138 + $22 = $160 | ~$40 LLM | ~$200 |
| Vapi | 2000 × $0.05 = $100 platform | ~$80 LLM + ~$40 STT + ~$80 TTS = $200 | ~$300 |
| Agora | 300 free + 1700 × $0.10 = $170 | Passthrough LLM | ~$210+ |
| TRTC | 990 free + 1010 × $0.10 = $101 | BYO LLM ~$40 | ~$141 |
Scenario B — Education/practice companion: 10,000 Conv AI min/month
| Vendor | Hosting | LLM/TTS/STT | Total |
|---|---|---|---|
| ElevenLabs Pro ($99) | 1,238 free + (10000-1238) × $0.08 = $701 + $99 = $800 | ~$200 LLM | ~$1,000 |
| Vapi | 10000 × $0.05 = $500 | ~$400 LLM + ~$200 STT + ~$400 TTS = $1,000 | ~$1,500 |
| Agora | 300 trial + 9700 × $0.10 = $970 | Passthrough LLM | ~$1,170 |
| TRTC | 990 free + 9010 × $0.10 = $901 | ~$200 LLM | ~$1,101 |
Scenario C — Enterprise call center: 100,000 Conv AI min/month
| Vendor | Hosting | Notes |
|---|---|---|
| ElevenLabs Business ($990) | 12,375 free + 87,625 × $0.08 = $7,010 + $990 = $8,000 | Plus LLM passthrough ~$2,000 |
| Vapi | 100k × $0.05 = $5,000 | Plus ~$10,000 LLM/STT/TTS |
| Agora | 100k × $0.10 = $10,000 | Plus passthrough |
| TRTC | 990 free + 99,010 × $0.10 = $9,901 | Plus BYO LLM |
| Enterprise custom (all vendors) | — | Expect 30–50% discount on volumes above 50k min/mo |
Read: At MVP scale, TRTC and ElevenLabs Creator are comparable. At education/practice scale (10k min/mo), Vapi's BYO model surfaces its hidden passthrough costs. At enterprise scale, all vendors negotiate — list prices become a starting point, not the final quote.
Free Conversational AI — what "free" actually means by vendor
The phrase "free Conversational AI" gets used a lot in 2026 marketing. Here's the decode:
| Vendor's "free" | What it really means |
|---|---|
| ElevenLabs Free tier | 15 Conv AI min/month. Enough for demos, not for production. |
| Vapi free credit | $10 signup credit ≈ 65 minutes on budget stack. One-time. |
| Agora free Conv AI | 300-minute trial. One-time. Does not renew. |
| "Free open-source voice agent" (various) | Framework is free; LLM/STT/TTS/hosting bills meter regardless. |
| TRTC free Conv AI | ~1,000 min/mo recurring across every plan, drawn from the unified 10k free-minute pool. |
If you're evaluating "free Conversational AI" offers in 2026, the question to ask is: "Does the free bucket renew monthly, or is it a one-time trial?" That single filter eliminates most of the competition.
Conversational AI pricing by industry / use case
Long-tail searches in this category often include an industry modifier. Here's how pricing typically lands:
Conversational AI pricing for home services (plumbing, HVAC, cleaning)
Typical call duration: 3–5 minutes. Typical monthly volume: 1,000–5,000 calls. At $0.10–$0.20/min all-in, monthly Conv AI spend runs $300–$3,000. Vendors actively marketing into this segment: GoHighLevel (Voice AI inside their agency CRM), Retell AI, Synthflow. TRTC and ElevenLabs work here but aren't specifically marketed at the home-services niche.
Conversational AI pricing for lending / fintech
Compliance requirements drive vendor choice more than price. Typical volume: 10,000–100,000 min/month. Expect enterprise-tier negotiated pricing. SAP Conversational AI and enterprise Hume AI deployments are common; TRTC is increasingly relevant for teams that want China Mainland compliance alongside global coverage.
Conversational AI pricing for education / language practice
High volume, cost-sensitive. 1,000–20,000 Conv AI minutes per user per month is common for daily-use practice products. Per-minute math matters most here. TRTC's monthly free pool is structurally useful — it effectively subsidizes daily demo users.
Vendor spotlights (with long-tail keyword coverage)
ElevenLabs Conversational AI pricing per minute
$0.08/min hosting across Free through Business tiers. Burst rate $0.16/min for 3× concurrency expansion. LLM tokens billed separately to your ElevenLabs credits. Telephony passed through at cost. Included minutes scale with plan (15 → 12,375/month).
Telnyx Conversational AI pricing per minute
Telnyx publishes Voice API rates publicly but Conversational AI / Voice AI pricing often requires an account or sales engagement. Verify current Telnyx Conv AI rates directly before publishing.
ElevenLabs Conversational AI pricing 2025 vs 2026
The 2025 rate card showed tiered hosting ($0.08–$0.30/min). The 2026 public page on elevenlabs.io/pricing/agents shows a flat $0.08/min hosting rate across all tiers. If your Q4-2025 estimates used tiered rates, refresh them.
SAP Conversational AI pricing
Enterprise custom. Typically bundled into broader SAP CX or CAI Recast agreements. Not listed publicly per-minute.
Hume AI EV2 Beta conversational ai pricing
EV2 Beta access and pricing have varied during the beta window. Confirm with Hume sales for current beta rates and production transition path.
GoHighLevel Voice AI pricing
Billed via your HighLevel subaccount. Rates are shown in the billing console after login. Public pages don't disclose the per-minute rate; published third-party estimates suggest ~$0.13–$0.16/min including telephony and LLM. Agency/SaaS Pro ($497/mo) enables resellable Voice AI.
FAQ
How much does Conversational AI cost per minute in 2026? The hosting layer alone ranges from $0.05/min (Vapi platform fee) to $0.10/min (Agora, TRTC) to $0.08/min flat (ElevenLabs Agents). With LLM tokens, STT, TTS, and telephony added, real all-in rates typically fall between $0.12 and $0.42/min depending on the stack.
Is there a free Conversational AI platform? TRTC offers the most generous recurring free bucket — approximately 1,000 minutes per month on every plan, drawn from a unified 10,000-minute monthly pool that also covers TTS, STT, and real-time translation. ElevenLabs Free gives 15 minutes/month. Vapi gives $10 signup credit (one-time). Agora offers a 300-minute one-time trial.
How is ElevenLabs Conversational AI priced? $0.08/min hosting, flat across all paid tiers. $0.16/min burst rate during temporary concurrency expansion. LLM tokens and telephony billed separately from the ElevenLabs credits pool. Included minutes scale with plan (Free 15 → Business 12,375).
Does the minute-based price include LLM tokens? Usually no. ElevenLabs, Vapi, and TRTC all require you to bring (or pass through) your own LLM. Agora's $0.10/min Conv AI is the hosting rate; LLM is passthrough. SAP and Hume tend to bundle (at enterprise pricing). Always ask your sales contact whether LLM is in or out of the quote.
What's cheaper: Vapi, Telnyx, or ElevenLabs? At the platform-fee layer, Vapi ($0.05) is cheapest. Once you add LLM/STT/TTS/telephony passthrough, ElevenLabs's $0.08 flat rate with built-in TTS/STT often comes out comparable to Vapi's $0.115+ all-in. Telnyx is difficult to compare on public data; get a quote. For teams that want predictable math without stack management, ElevenLabs or TRTC are simpler.
How do I calculate Conversational AI TCO? Build a spreadsheet with five lines: (1) hosting rate × minutes, (2) LLM tokens × expected turns per call × turns rate, (3) TTS characters × expected output length × TTS rate, (4) STT minutes × rate, (5) telephony minutes × rate. Multiply each by expected monthly volume. Compare totals, not hosting rates.
Can I mix providers (STT + LLM + TTS)? Yes. Vapi, LiveKit Agents, and Pipecat (Daily.co) all treat LLM/STT/TTS as swappable passthrough. TRTC bundles STT/TTS but supports BYO LLM. ElevenLabs Agents bundles TTS and lets you pick LLM and STT. Choose based on whether you value "one bill, one pool" (TRTC, ElevenLabs) or "cheapest possible per component" (Vapi, LiveKit).