Tasks and inputs
Eleven v4 Turbo is listed for Text → Speech, Text-to-speech. It accepts text and produces audio. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Nov 17, 2026
Eleven v4 Turbo is ElevenLabs' low-latency v4 speech model for voice agents, with about 100 ms median inference, 90+ languages and half v4's price.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: About $0.04 per 1K characters ($40 per 1M characters). Indicative; confirm on your ElevenLabs plan.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
A voice agent lives or dies in the gap after the user stops talking. Eleven v4 Turbo is built for that gap. It is the real-time member of ElevenLabs' v4 family, released on 28 September 2026, with median inference of about 100 ms and around 150 ms median time to first speech over WebSocket streaming (vendor figures from September 2026).
Those two numbers measure different things, and the larger one is often quoted as the headline. 100 ms is model compute. 150 ms is what a client sees before audio starts, network excluded. Your own figure, measured from your region, will be higher than both.
Two changes since launch tell you how ElevenLabs sees it:
eleven_flash_v2.If you build on ElevenAgents, you may already be using it.
| v4 Turbo | v4 | Flash v2.5 | |
|---|---|---|---|
| Model ID | eleven_v4_turbo |
eleven_v4 |
eleven_flash_v2_5 |
| Median latency | about 100 ms | not latency-tuned | about 100 ms |
| Languages | 90+ | 90+ | 32 |
| Transport | Text to Dialogue WebSocket | Text to Dialogue API | - |
| Audio tags | Yes | Yes | Not listed |
| Indicative price | $0.04 / 1K chars | $0.08 / 1K chars | lower |
Versus Flash v2.5, the gain is not raw speed; both sit around 100 ms. It is language breadth and expressiveness: audio tags and high-fidelity voice cloning come along. Versus the flagship, you trade some performance depth for latency and a 50% lower rate.
Each open Text to Dialogue WebSocket connection holds one session, and plans cap concurrent sessions (from 14 on Free up to 105 on Scale and Business). Idle connections close after 20 seconds unless you send keep_alive messages. Size your plan to peak concurrent calls, not daily volume.
Instrument the client, not the server. Log the timestamp when the user's turn ends and when the first audio byte plays, for a few hundred real turns, and compare Turbo with whatever you run today. Then listen to a sample of those turns with audio tags in the text; expressiveness is the reason to choose Turbo over Flash. Price the result against MiniMax Speech 2.8 TURBO and Voxtral TTS, or scan other text-to-speech models.
Compare the model on the work you need to complete.
Eleven v4 Turbo is listed for Text → Speech, Text-to-speech. It accepts text and produces audio. Use the publisher’s input format and preprocessing for your chosen deployment.
The listed route has starting rates of $40.00 input per million tokens. Compare pronunciation, voice consistency and time to the first playable audio. Check the service’s billing unit for the actual script.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
inclusionAI
Ming-3.0-tts-0.5B is a compact open text-to-speech model from inclusionAI with voice design, zero-shot cloning and fine control over delivery.
Sarvam
Bulbul V4 Flash is Sarvam's fast text-to-speech model for 11 Indian locales, with sub-250 ms first audio at a third less than Bulbul V3.
ElevenLabs
Eleven v4 is ElevenLabs' flagship text-to-speech model for audiobooks, dubbing and emotional dialogue, with 90+ languages and 10,000 characters per request.