GA Ultravox (formerly Fixie.ai)

Ultravox Realtime

Hosted voice agents built on Ultravox's audio-native model, which listens to speech directly (no separate transcription) and answers through TTS voices, at a flat $0.05 per minute. A good default for cost-sensitive phone and web agents; the speech-understanding model weights are open.

Est. per minute$0.05 - 0.055
1 high-severity warning

Overview

Best for: Budget phone agents and web voice assistants that need decent reasoning at a flat, low per-minute price.

At a glance

Flat $/min$0.05
Free tierYes
Native S2SNo
ToolsYes
CloningYes
Concurrency5
WebRTCYes
WebSocketYes
Phone / SIPYes
Open weightsYes

Audio-native understanding model (no ASR stage) with separate TTS output, so not a single speech-to-speech model. Open weights are the MIT audio adapter only. maxDuration defaults to 1 hour, no documented maximum. Concurrency hard cap 5 on PAYG; Pro has no hard cap. 30 free call minutes.

Audio in

Raw PCM s16le for server WebSocket (sample rate set in the call medium)

Audio out

PCM for server WebSocket; WebRTC handles codecs automatically

Languages

Multilingual; no list on the pages read.

Voices

Ultravox voice library plus voice clones (1 on PAYG, 5 custom voices on Pro).

Latency

Vendor says audio-native processing is faster and robust to transcription errors; no figure on the pages read.

Regions

Not stated on the pages read.

Compliance

Not stated on the pricing or FAQ pages (a trust portal is linked).

Features

  • audio-native understanding (no ASR stage)
  • tool calling
  • RAG corpora (2 PAYG, 20 Pro)
  • outbound call scheduler (Pro)
  • call recordings and transcripts
  • playground (free)

Pricing

WhatPriceUnitNotes
Calls (Pay As You Go and Pro)$0.05per minuteRounded to 6-second 'deciminutes' ($0.005 each); no surge pricing.
SIP$0.005 PAYG / $0.0048 Proper minute
Text/thread usage$2.00 in (uncached) / $15.00 outper 1M tokensCached input not charged; table layout ambiguous about which plans.
Pro plan$100per monthPlatform fee; minutes billed on top; no hard concurrency cap.
How the per-minute estimate was worked out

$0.05 call rate plus $0.005 SIP when using SIP; LLM and TTS included.

Audio token rate

Not token-billed for calls.

Free tier: 30 free call minutes; playground calls free.

Source: ultravox.ai

Setup

  1. Create an account and API key in the Ultravox console.
  2. POST /api/calls with systemPrompt, model, voice, medium and maxDuration.
  3. Join the returned joinUrl from a browser (WebRTC client SDK) or bridge it to telephony (Twilio, SIP, server WebSocket).

Endpoint

POST https://api.ultravox.ai/api/calls (returns joinUrl)

Authentication

X-API-Key header

Quick start javascript

// Server side: create a call (Node 18+, global fetch)
const res = await fetch("https://api.ultravox.ai/api/calls", {
  method: "POST",
  headers: { "X-API-Key": process.env.ULTRAVOX_API_KEY, "Content-Type": "application/json" },
  body: JSON.stringify({
    systemPrompt: "You are a friendly receptionist for Acme Dental.",
    model: "ultravox-v0.7",
    maxDuration: "600s",          // default is 1 hour; cap it to control spend
    // medium: { serverWebSocket: { inputSampleRate: 16000, outputSampleRate: 16000 } },
  }),
});
const { joinUrl } = await res.json();

// Browser side (WebRTC is the default medium): npm i ultravox-client
// import { UltravoxSession } from "ultravox-client";
// const session = new UltravoxSession();
// session.joinCall(joinUrl);
// ... later: session.leaveCall();

Written from the current docs. Check the vendor's SDK version before you ship.

Warnings

5-call concurrency on PAYG

Pay As You Go has a hard cap of 5 concurrent calls; anything beyond a pilot needs Pro ($100/month) or Enterprise.

Model migrations are forced

Qwen3-based models were retired with about two weeks' notice in December 2025 and the default moved to v0.7. Pin the model string and watch announcements.

Default call length is 1 hour

maxDuration defaults to 3600 s; a stuck call can run an hour at $0.05/min. Set a shorter maxDuration.

Open weights are not the whole product

The MIT-licensed v0.7 checkpoint is an audio adapter on GLM-4.6 (a very large MoE model) and outputs text, not speech; self-hosting needs GLM-4.6 serving plus your own TTS.

Deletion is permanent

Call data is retained until you delete it, and deletion cannot be undone; set a retention process for compliance.

Token charges for text threads

Text/thread usage is billed per token ($2 in / $15 out per 1M), separate from call minutes.

Plus 14 warnings that apply to all voice-to-voice APIs. See category warnings.

Limits

  • Pay As You Go: hard cap of 5 concurrent calls
  • maxDuration defaults to 1 hour per call
  • Invoiced monthly in arrears; early invoice at $10 (PAYG) or $100 (Pro) usage thresholds

Models and products

NameStatusNotes
ultravox-v0.7GADefault since Dec 2025; built on GLM 4.6. Open weights (MIT) for the audio adapter.
ultravox-v0.6 / fixie-ai/ultravox-llama3.3-70bGALegacy (Llama 3.3 70B); must be set explicitly.
Qwen3-based UltravoxDeprecatedPhased out December 2025.

Docs and sources

Docs

Sources used

Not fully verified

Language list, latency, compliance certifications, exact browser SDK method names (ultravox-client UltravoxSession.joinCall).

Similar voice-to-voice APIs

Spotted a wrong price or a dead link?