74 / 100

Retell AI pricing teardown

https://www.retellai.com/pricing

Retell uses a genuinely usage-based pay-as-you-go model with a built-in cost estimator and very granular component pricing, which gives developers and SMBs strong transparency and a zero-commitment entry. The tradeoff is a binary Pay-as-you-go vs. Enterprise split with no mid-tier, and a very dense, catalog-like price breakdown (dozens of LLM SKUs) that makes the true per-minute cost hard to predict at a glance.

Tier structure

Pay as you go

$0.07–$0.31/min (voice), $0.002+/msg (chat)

Enterprise

Custom pricing

Value metric

usage-based per minute (voice) and per message (chat), with per-component pricing

How limits scale

Concurrent calls 20 included (+$8/mo each) → No cap / custom (50+)
Server Shared → Dedicated stable server (add-on)
Support Community and email → 24/7 with dedicated portal + named account manager
Starting credits $10 free credits → Custom

Escalation logic

A two-step ladder: self-serve metered usage with full platform access, then a custom-quoted Enterprise plan for compliance, dedicated infrastructure and support. There is no middle plan, so the upgrade trigger is security/compliance needs or very high concurrency rather than feature gating.

Psychological anchors

  • Free credits as entry anchor — '$10 in free credits' and 'starts at $0' lower the barrier to trying the platform, with no contract required.
  • Interactive cost estimator — A volume slider plus LLM/TTS/telephony/add-on selectors produces a live cost-per-minute and monthly total, turning an abstract rate card into a concrete number.
  • Dark Enterprise card as contrast/anchor — The Enterprise plan is visually inverted (dark card) and framed around reliability and compliance, making the pay-as-you-go card read as the lightweight, accessible option.
  • 'Recommended' labels on models — GPT 5.6 Terra, Claude 5 Sonnet, GPT 5.5 and GPT 5.4 are flagged Recommended, steering users through a huge LLM menu (a decoy/default effect).
  • Price range framing — '$0.07–$0.31/min' shows a low floor to attract users while letting the configuration determine the real cost.
  • Anti-contract positioning — 'True pay as you go' and 'most voice AI platforms require annual contracts' position Retell against competitors' lock-in.
  • Volume-triggered enterprise nudge — The estimator surfaces 'Enterprise pricing applies at this volume — Get a volume quote' to route large buyers to sales.

What this page is optimizing for

A hybrid: self-serve developer/SMB conversion (public per-minute prices, free credits, 'Sign up for free' CTA, no gating of features) with an enterprise lead-gen path ('Book a demo', custom pricing, volume-quote nudge). The emphasis is on low-friction acquisition and usage growth, with enterprise upsell driven by compliance and concurrency.

Red flags

  • No mid-tier plan: the jump from pure metered usage to fully custom Enterprise leaves growing teams with no predictable, packaged option.
  • The headline range ($0.07–$0.31/min) is wide, and the true cost depends on stacking LLM + voice infra + TTS + telephony + add-ons, so budgeting is hard without the calculator.
  • The detailed component tables list 40+ LLM SKUs and many TTS options with a Standard vs. Fast tier, causing choice overload and making comparisons tedious.
  • Several costs surface only below the fold (concurrency at $8/mo each beyond 20, phone numbers $2/mo, verified numbers $10/mo, billing during silence/hold), and a few hidden-feeling add-ons (PII removal, guardrails, denoising) each add per-minute fees.
  • No annual or committed-use discount is shown, so high-volume buyers see no visible price break before talking to sales.
  • Comparison table shows HIPAA/BAA, SSO and Custom MSA/DPA as Enterprise-only while the benefits section claims 'no feature gating', which creates messaging tension.
  • Little social proof or trust signaling on the page itself, with no logos or testimonials visible near the pricing cards.

Best-practices scorecard

What works · 5

  • Visible prices — Per-minute, per-message and per-component prices are fully public; only Enterprise is custom.
  • Value metric matches usage — Per-minute and per-message billing, tracked to the second and charged only for connected calls, tracks closely with value delivered.
  • Tier differences are scannable — The Compare Plans table with checkmarks cleanly separates Pay-as-you-go from Enterprise across Voice AI, Infrastructure, Compliance and Support.
  • FAQ or objection handling — FAQ covers minute calculation, transfers, silence billing, failed calls, invoices and concurrency limits, addressing the main billing surprises.
  • Clear CTAs per tier — 'Sign up for free' on Pay-as-you-go and 'Book a demo' on Enterprise match each tier's buying motion.

Half measures · 2

  • Clear recommended tier — Only two plans and no 'most popular' badge; the Pay-as-you-go card is the implicit default and the 'Recommended' tags apply to LLM models rather than plans.
  • Trust signals present — SOC 2, HIPAA-ready and SSO are mentioned and a $100M milestone banner appears, but there are no logos, reviews or testimonials near the pricing.

What's missing · 1

  • Annual discount offered — No monthly/annual toggle or committed-use discount is shown; it's pure usage billing.