Eleven v4 Turbo 上线 ElevenAgents

Blog; Product

ListenListen to this article

On this page

Introduction; Expressive speech at conversation speed; In whatever language your customers speak; Built for your industry; One co-optimized stack; Every conversation, at its best

Yesterday we launched Eleven v4, our most emotive text-to-speech model yet, and its low-latency variant, Eleven v4 Turbo. Eleven v4 Turbo is now available in ElevenAgents, powering more personal, empathetic customer experiences across industries.

For your customers, that means an agent that sounds right for the moment - apologetic about a missed delivery, brisk when someone just wants a confirmation - in a voice that stays the same across many calls.

Expressive speech at conversation speed

Ranked #1 by Artificial Analysis1, and preferred by ~75% of listeners in blind head-to-head tests over competing models2, Eleven v4 was designed to interpret tone, pacing, emotion, personality, and context.

Eleven v4 Turbo in ElevenAgents brings this wide expressive range, natural tone, and pacing to live conversations, with a median inference latency of ~100 ms and high reliability on long calls.

The agent understands conversations as they unfold and adjusts mid-call - reassuring when a caller is frustrated, clearer when they are confused - across support, sales, scheduling, and more.

In whatever language your customers speak

Eleven v4 Turbo supports over 90 languages, with the biggest improvements in Japanese, Brazilian Portuguese, Mandarin, and Cantonese. When a caller switches languages mid-call, the agent switches with them - keeping recognizably the same voice, or switching voices for native-level accuracy. And with the pronunciation dictionary, you add a brand, drug, or place name once, and your agents an say them correctly across languages.

Built for your industry

With Eleven v4 Turbo, agents rise to the moments that make or break customer trust. Pick your industry to hear our agents navigating high-stakes scenarios with empathy.

Hear an agent booking a sensitive procedure for a patient worried about cost - with a tone that fits the moment.

Hear an agent resolving a damaged order right before a customer's trip - and helping them size their next purchase.

Hear an agent gracefully handling a dispute from a worried customer with an unrecognized transaction.

Hear an agent helping a resident sort out a parking citation - and proactively surfacing the permit that prevents the next one.

Hear an agent running live diagnostics for a frustrated customer who needs to get back online.

One co-optimized stack

In ElevenAgents, Eleven v4 Turbo works alongside our transcription and turn-taking models. One co-optimized stack means agents that respond faster, interrupt gracefully and improve automatically as new models launch.

Every conversation, at its best

Eleven v4 Turbo makes natural, expressive conversations the default in ElevenAgents.

Learn more about Eleven V4 in our blog, or request a demo below.

Get an ElevenAgents demo

Artificial Analysis, Provider Voice Arena Leaderboard, Sept 2026; Based on blind head-to-head user preference testing against Cartesia Sonic 3.6, Inworld TTS-2, Google Gemini 3.8 Flash-Lite TTS, and Google Gemini 3.8 Flash TTS, September 2026. For each pair, graders heard the same line from Eleven v4 and one competitor, presented blind, and judged which was more expressive and which sounded more natural; ties counted as half

Research - Speech to Text and Text to Speech