Eleven v4: ElevenLabs' New TTS Model, Pricing and Specs

Eleven v4 is ElevenLabs' most expressive text-to-speech model, released September 28, 2026, with a lower-latency sibling, Eleven v4 Turbo. As of October 17, 2026, the API charges $0.08 per 1,000 characters for v4 and $0.04 for v4 Turbo, and the two models hold the top two places on the Artificial Analysis voice leaderboard.

This guide covers what v4 is, what changed from earlier ElevenLabs models, what it costs on the API and on plans, and how it ranks. ElevenLabs' figures are vendor-reported. Rankings come from Artificial Analysis (AA) and reflect a snapshot taken the same day.

Eleven v4 and Eleven v4 Turbo at a glance

| Feature | Eleven v4 | Eleven v4 Turbo | |---|---|---| | Model ID | eleven_v4 | eleven_v4_turbo | | Designed for | Highest quality output | Real-time voice agents | | Access | Text to Dialogue API | Text to Dialogue WebSocket | | Input limit | 10,000 characters (about 10 minutes of audio) | Not stated in this guide | | Latency (vendor claim) | Not stated | About 100ms median inference, about 150ms median time to first speech |

ElevenLabs measured the Turbo latency in September 2026 with network latency removed, using WebSocket streaming. It compares this with 262 to 814ms for competitors in its chart. Those are ElevenLabs' numbers under its own test conditions.

Source: ElevenLabs, Introducing Eleven v4 and ElevenLabs model docs.

What is new in Eleven v4

ElevenLabs describes v4 as its most expressive model, built on a new architecture. These points come from the launch post and are vendor claims:

It interprets tone, pacing, emotion, character and context.
Multi-speaker dialogue sounds more natural.
Audio tags such as [laughs], [light rain] and [phone buzzing] are followed more accurately.
You can describe delivery in plain language, as well as with tags.
Phoneme support for IPA is better.
It supports 90+ languages, and the voice keeps its identity across languages with stronger accent adherence.

The language count applies to the model. ElevenLabs' pricing page says "74 on every plan," so do not treat either figure as a single number for "ElevenLabs."

For comparison, the model docs list these other ElevenLabs models:

| Model | Languages | Character limit | Notes from the docs | |---|---|---|---| | eleven_v3 | 70+ | 5,000 | Earlier expressive model | | eleven_multilingual_v2 | 29 | 10,000 | Earlier multilingual model | | eleven_flash_v2_5 | 32 | 40,000 | About 75ms latency, about 50% lower API price per character | | eleven_flash_v2 | English only | Not stated in this guide | Flash model |

The docs also mark eleven_turbo_v2_5 and eleven_turbo_v2 as deprecated.

Voice cloning in Eleven v4

ElevenLabs says Instant Voice Clones need about 10 seconds of audio. v4 also supports Professional Voice Clones, and ElevenLabs reports better speaker similarity and consistency in long-form audio. These are vendor claims. Plan access matters here: Instant Voice Cloning starts on the Starter plan, and Professional Voice Cloning starts on Creator.

Eleven v4 pricing

API prices are per 1,000 characters, in USD. The per-1M figures are derived by multiplying by 1,000 and match AA's listed prices.

| Model | Per 1,000 characters | Per 1M characters (derived) | |---|---|---| | Eleven v4 | $0.08 | $80 | | Eleven v4 Turbo | $0.04 | $40 | | Multilingual v2 | $0.08 | $80 |

Source: ElevenLabs API pricing.

Plan prices, billed monthly, as of October 17, 2026:

| Plan | Monthly price | Included credits | |---|---|---| | Free | $0 | 10,000 | | Starter | $6 | 30,000 | | Creator | $22 | 121,000 | | Pro | $99 | 600,000 | | Scale | $299 | 1.8 million | | Business | $990 | 6 million | | Enterprise | Custom | Custom |

Annual billing is priced at 10 months for 12. The commercial license starts at Starter. ElevenLabs does not publish the credit cost per character for v3 and v4 on the pricing page, so this guide does not convert credits into minutes for v4. Check the current multiplier before you budget.

Source: ElevenLabs pricing.

Where Eleven v4 ranks

The Artificial Analysis Provider Voice Arena ranks text-to-speech models by Elo, which comes from user votes. The table shows the top five on October 17, 2026. The price is what AA says it costs to generate 1 million characters on the creator's API with default settings.

| Rank | Model | Elo | Price per 1M characters | |---|---|---|---| | 1 | Eleven v4 Turbo | 1330 | $40.00 | | 2 | Eleven v4 | 1329 | $80.00 | | 3 | Qwen-Audio-3.1-TTS-Plus (Alibaba) | 1296 | $19.30 | | 4 | Cartesia Sonic 3.6 | 1285 | $49.00 | | 5 | Gemini 3.8 Flash TTS | 1275 | $16.50 |

Source: Artificial Analysis text-to-speech leaderboard.

Two points matter here. The Elo gap between v4 Turbo and v4 is one point, so there is no meaningful difference between them on this measure. And Elo reflects listener preference in one test, not every use case.

Eleven v4 vs other new voice models

Other vendors have released voice models close to the same period. Each row uses that vendor's published pricing.

| Model | Vendor and status | Published price | |---|---|---| | Gemini 3.8 Flash TTS | Google, generally available September 22, 2026 | $0.50 per 1M text input tokens and $9.00 per 1M audio output tokens, through December 31, 2026 | | GPT-Live 1 | OpenAI, generally available September 10, 2026 | $0.05 per minute for voice sessions, billed per second; backend model and tools billed separately | | grok-voice-3.0 | xAI | $0.08 per minute speech to speech; $15.00 per 1M characters text to speech |

Google says the Gemini 3.8 Flash TTS audio output price doubles from January 1, 2027. Gemini 3.8 Flash TTS supports over 130 languages, 30 prebuilt voices and up to two speakers per request. Source: Google's Gemini API pricing page.

ElevenLabs also reports blind listening results. It says v4 won 65% to 81% of head-to-head comparisons against Cartesia Sonic 3.6, Inworld TTS-2 and two Gemini 3.8 Flash TTS models, with ties counted as half. It says about 75% of listeners preferred v4 overall. These are vendor charts, not independent tests.

What the numbers do not tell you

Several caveats apply to the figures above:

Vendor charts show latency and preference results measured under the vendor's own setup.
Price units differ. Some models bill per token and others per character or per minute. Compare them only on your own volume.
Language counts differ by source. The model page says 90+ for v4 and the pricing page says 74 on every plan.
Elo is not quality for your content. It measures votes in one leaderboard.

How to choose a voice model

Decide whether you need real time. Voice agents need low latency, so start with the Turbo tier and a streaming connection.
Check the length limit. Eleven v4 accepts up to 10,000 characters per request in the API.
Check your languages against the model's list, not a single headline number.
Confirm the plan tier for cloning and commercial use.
Price it at your volume. Multiply your monthly characters by the per-character rate.
Listen to your own content. Leaderboards and vendor charts are not a test of your script.

Where OndaLabs fits

OndaLabs is an AI voice studio. Its text to speech includes 30 prebuilt voices, usable in every supported language, and 37 language pages. Its two-speaker dialogue feature covers scripts with two voices. Custom voices can be cloned from a recording you have permission to use, or designed from a text description, at a cost of 1,500 credits. Onda Lite costs 1 credit per character and is available on all plans. Onda Flash costs 1.5 credits per character and is available on paid plans. The Free plan includes 2,000 credits per month, and Starter costs $9 for 170,000 credits, with commercial use included.

As a derived comparison, using published plan prices: Onda Lite on Starter comes to about $52.94 per 1M characters, and Onda Flash to about $79.41. These are different products on different billing, so they are not like-for-like with the Eleven v4 API. See OndaLabs pricing for current plans.

For the text models that often pair with voice, see The best AI models right now: October 2026 roundup. Text models like GPT-6 Astra return text only, so a voice model is a separate step.

Frequently asked questions

What is Eleven v4?

Eleven v4 is ElevenLabs' newest text-to-speech model, released September 28, 2026. ElevenLabs describes it as its most expressive model, with better tone and emotion control, audio tags and multi-speaker dialogue. It is available through the ElevenLabs API, ElevenAgents and ElevenCreative.

What is the difference between Eleven v4 and Eleven v4 Turbo?

Eleven v4 is built for the highest quality and is accessed through the Text to Dialogue API, with a 10,000-character limit per request. Eleven v4 Turbo is built for real-time voice agents and uses a WebSocket connection. Turbo also costs less per character, at $0.04 against $0.08 per 1,000 characters.

How much does Eleven v4 cost?

On the API, Eleven v4 costs $0.08 per 1,000 characters, which is $80 per million characters. Plan prices start at $0 on Free, $6 on Starter and $22 on Creator, billed monthly. The pricing page does not publish a credit multiplier for v4, so check it before you budget.

How many languages does Eleven v4 support?

ElevenLabs says Eleven v4 supports 90+ languages, and that voices keep their identity across languages. The pricing page says 74 languages are available on every plan. Check the language list for your target language before you commit.

Does Eleven v4 support voice cloning?

Yes. ElevenLabs says Instant Voice Clones need about 10 seconds of audio, and Professional Voice Clones are supported in v4. Instant Voice Cloning starts on the Starter plan and Professional Voice Cloning starts on Creator. Use voices only from recordings you have permission to use.

Is Eleven v4 the best text-to-speech model right now?

It is ranked first and second on the Artificial Analysis Provider Voice Arena as of October 17, 2026, with v4 Turbo at 1330 Elo and v4 at 1329. That ranking is based on user votes and is a snapshot. Test the models on your own content, because preference in one test does not decide every use case.