ElevenLabs Adds Eleven v4 Turbo to ElevenAgents for Low-Latency Voice Interactions

ElevenLabs has integrated Eleven v4 Turbo into ElevenAgents, bringing its fastest and most emotive text-to-speech capabilities to live customer interactions. The model is designed to interpret tone, pacing, and context, allowing agents to adjust their delivery mid-call—such as sounding apologetic for a missed delivery or brisk for a simple confirmation.

Performance and Expressiveness

Eleven v4 Turbo achieves a median inference latency of ~100 ms, ensuring high reliability on long calls while maintaining natural pacing. According to Artificial Analysis, Eleven v4 ranks #1 on the Provider Voice Arena Leaderboard. In blind head-to-head tests against competitors like Cartesia Sonic 3.6, Inworld TTS-2, and Google Gemini 3.8 Flash-Lite TTS, listeners preferred Eleven v4 approximately 75% of the time for expressiveness and naturalness.

Multilingual Capabilities

The model supports over 90 languages, with significant improvements noted in Japanese, Brazilian Portuguese, Mandarin, and Cantonese. Agents can switch languages mid-call while maintaining a consistent voice identity or switching to a native-level voice for accuracy. A pronunciation dictionary allows developers to define brand, drug, or place names once, ensuring correct pronunciation across all supported languages.

Integrated Stack

Eleven v4 Turbo operates within a co-optimized stack alongside ElevenLabs' transcription and turn-taking models. This integration enables faster responses and graceful interruptions, with the system automatically improving as new models are released. The update targets high-stakes scenarios in healthcare, retail, financial services, government, and telecommunications.