ElevenLabs launches Eleven v4 and v4 Turbo, with 90+ languages

ElevenLabs released Eleven v4, its new text-to-speech model, on September 28, 2026, together with a low-latency version, Eleven v4 Turbo. Both support more than 90 languages and are live in ElevenAgents, ElevenCreative and the ElevenLabs API. ElevenLabs says a free account can use them.

What’s new

• Two models: eleven_v4 for expressive speech and eleven_v4_turbo for voice agents and other real-time use.
• Languages: more than 90, for both models.
• Voice cloning: ElevenLabs says Instant Voice Clones now capture a voice from 10 seconds of audio. Professional Voice Clones also work with v4.
• Speed: ElevenLabs gives Turbo a median inference latency of about 100 ms and a median time to first speech of about 150 ms, in its own tests.
• Length: the docs list a limit of 10,000 characters per request for Eleven v4, about 10 minutes of audio.

ElevenLabs also says v4 ranked first in the Artificial Analysis Provider Voice Arena in September 2026. That ranking and the latency figures come from ElevenLabs’ own post.

Older models

The docs now call Eleven v3 and Multilingual v2 “previous generation” models, but they are still listed. Turbo v2.5 and Turbo v2 are marked deprecated; ElevenLabs recommends its Flash models instead. Flash v2.5 is still offered, with a 40,000-character limit and 32 languages.

Price

Neither the launch post nor the pricing page lists a separate credit rate for v4. The plans on the pricing page on October 18, 2026:

Plan	Monthly	Credits a month	Notes
Free	$0	10,000	3 Studio projects
Starter	$6	30,000	Adds a commercial license
Creator	$22	121,000	Professional voice cloning
Pro	$99	600,000	44.1 kHz PCM audio via API

Paying yearly costs ten times the monthly price.

Already in other tools

Runway added ElevenLabs V4 for audio generation on September 29, 2026. Descript’s AI Speech switched to ElevenLabs v4 in its October 12, 2026 release.

If your app still calls Turbo v2.5, it is on a deprecated model: test v4 Turbo or Flash v2.5 and compare the credits each request uses in your usage page.