Laboratory
Entity that develops or publishes the model and maintains its technical and safety documentation.
VERIFIED ORGANISATION
ElevenLabs
Explore the documented models, published evaluations and primary sources associated with this organisation.
01 / MODELS
Verified catalogue
EXPRESSIVE VOICE SYNTHESIS
Eleven v3
Expressive text-to-speech model with emotional tags, multi-speaker dialogue, and support for more than 70 languages.
18 SEP 2026↗ MULTILINGUAL TRANSCRIPTIONScribe v2
Speech recognition model for more than 90 languages, with diarization, word-level timestamps, entities, and audio labels.
18 SEP 2026↗ EXPRESSIVE VOICE SYNTHESISEleven v4
Expressive speech synthesis model for narration, characters, and dialogue in more than 90 languages.
29 SEP 2026↗ LOW-LATENCY SPEECH SYNTHESISEleven v4 Turbo
Eleven v4 variant optimized for voice agents and expressive real-time responses.
29 SEP 2026↗02 / EVALUATIONS
Published results
Figures are shown with the context reported by their source. A provider result is not an independent comparison and does not replace your own evaluation.
| Model | Benchmark | Result | Metric |
|---|---|---|---|
| Eleven v3 | Evaluación Eleven v3 | 70+ languages | Published coverage and expressiveness |
| Scribe v2 | Evaluación Scribe v2 | 90+ languages | Published ASR coverage |
03 / METHOD
How to read the directory
An organization, a product, and a model are not the same unit. Inferama separates them to avoid attributing capabilities or commercial terms to the wrong item.
Model
Identifiable version with limits, modalities, and behavior that may change between releases.
Access channel
API, application, associated cloud, or commercial plan; each channel may have different pricing, retention, and limits.
Source and date
Every claim must retain the official page consulted and the verification date.
ANALYSIS
Related analysis
Eleven v4 Turbo for Voice Agents: What We Know About Its Latency and What Still Needs Testing
ElevenLabs presents v4 Turbo as a low-latency variant and publishes median inference latency of around 100 ms. That figure is not the response time of a complete agent: measurement details, access conditions, applicable pricing, and model-specific security evidence are still missing.
02 Oct 2026 ↗
Eleven v4 Under Scrutiny: Documented Controls and Open Questions
ElevenLabs describes Eleven v4 as its most expressive text-to-speech model, with greater control over voice performance. That characterization comes from the provider: by itself, it neither quantifies quality nor demonstrates consistent results. Here is what the available sources support—and what a team would still need to verify.
01 Oct 2026 ↗
Eleven v3 for Multilingual Localization: What Can Be Claimed and What Must Be Tested
The materials provided associate Eleven v3 with multilingual voice generation, but they are not enough to verify a specific language count or the quality of a localization. We review what the available sources support and suggest checks to run before adding it to production.
29 Sep 2026 ↗
Scribe v2 in speech benchmarks: what its scores compare—and what is still needed to reproduce them
A WER score alone does not describe a transcription system’s overall performance. We examine what can be verified about Scribe v2, how the corpus, normalization, and configuration affect results, and what information is needed to repeat a comparison.
28 Sep 2026 ↗
How to Evaluate Eleven v3: Audio Tags, Stability, and Voice Consistency
A protocol for checking, with controlled voices and scripts, whether a delivery cue changes an interpretation without harming text fidelity, vocal identity, or repeatability. The available documentation is not sufficient to confirm the specific Eleven v3 tags here, so the first step is to verify what the access method being tested supports.
28 Sep 2026 ↗
DeepSeek V4.1 Flash + Eleven v3 vs. full-duplex voice: how to compare the architectures
A modular chain built from speech recognition, DeepSeek V4.1 Flash, and Eleven v3 is not equivalent to a full-duplex voice model. This guide proposes comparing both systems end to end, without declaring a winner based on isolated specifications.
26 Sep 2026 ↗04 / SOURCES