September 28, 2026
Voice AI agents could respond faster: Eleven v4 Turbo promises 100 ms
On Sep 28, ElevenLabs launched Eleven v4 Turbo with a claimed median latency of around 100 ms for voice AI agents. Eleven v4 and Turbo are already available in ElevenCreative, ElevenAgents, and ElevenAPI. For the first two weeks, the API charges $22 per 1 million characters for Eleven v4 and $11 for the same volume of Turbo.
ElevenLabs
@elevenlabs
Introducing Eleven v4 and Eleven v4 Turbo — our fastest and most emotional voice models. Artificial Analysis ranked them #1.
· 21.6K views
ElevenLabs' official quickstart still lists model_id="eleven_v3". The new models already work in ElevenAPI, ElevenCreative and ElevenAgents. The landing page is at elevenlabs.io/v4.
Text delivery. v4 accepts inline tags for emotion, pacing, reactions, SFX and style. A text description of the performance works instead of tags. Instant Voice Clone creates a voice from 10 seconds of audio.
ElevenLabs promises more consistent voice preservation in long-form voiceovers and dialogues across more than 90 languages. Artificial Analysis ranked Eleven v4 first for speech synthesis quality, with a Quality Elo of 1319 based on Speech Arena user responses.
In ElevenCreative, v4 is free within double the monthly credit limit. ElevenAgents with v4 or Turbo costs 3.3¢ per minute at the promotional price.
API promotional pricing is available for the first two weeks after the Sep 28 launch.
