# Scribe v3 Live Benchmark Scores & Performance

> Scribe v3 Live has ElevenLabs launch evaluation results. Voice evidence is display-only and does not produce a text-model score.

Canonical page: https://benchlm.ai/models/scribe-v3-live

Last updated: November 11, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | ElevenLabs |
| Source Type | Proprietary |
| Reasoning Type | Non-Reasoning |
| Context Window | N/A |
| Official model card | [ElevenLabs model card](https://elevenlabs.io/blog/scribe-v3) |
| Overall Score | Not scored (voice evidence only) |
| Overall Rank | Unranked |

## Family & Coverage

- Family: Scribe v3 Live
- Variant: base
- Benchmarks covered: 0 of 815
- Related earlier model: [Scribe v2.2 Realtime](/models/scribe-v2-2-realtime)
- Coverage note: Speech and voice results appear below; no weighted text-model benchmark rows are stored.

## Speech and voice evidence

[Scribe v3 launch evaluation](/voice-benchmarks/scribe-v3-launch-evaluation)

ElevenLabs ran every result itself, on an internal version of FLEURS and on its own unreleased dataset, so nothing here is independently reproducible yet. Each FLEURS language is scored on its own and then averaged, which ElevenLabs notes gives higher numbers than pooling clips, so these WERs are not comparable with pooled results elsewhere. The comparisons against other providers use only the languages both models support. The realtime diarization figure aligns each transcript to the reference first and is not comparable with the batch diarization figures. The launch post reports Scribe v3 Live on 102 FLEURS languages; the documentation reports 73. Semantic WER counts an error only when the meaning changes. These transcription results stay separate from BenchLM’s weighted text-model ranking.

| Metric | Result |
| --- | --- |
| Realtime diarization error rate, English (aligned) | 3.1% |
| FLEURS semantic WER vs Scribe v2 Realtime | 43% lower |
| FLEURS WER vs Scribe v2 Realtime | 31% lower |
| Earnings-22 WER | 3.91% |
| VoxPopuli WER | 1.71% |
| Price | $0.22 per hour |

ElevenLabs launch post and model documentation, read November 11, 2026. Provider-run; the diarization set is internal and unreleased (November 4, 2026), and transcripts were aligned to the reference before scoring, so the figure only compares within ElevenLabs’ chart. The post reports FLEURS over 102 languages; the documentation gives 44% lower semantic WER over 73 languages. Lower is better. Voice results are display-only.

- [Evaluation source](https://elevenlabs.io/blog/scribe-v3)

Released November 10, 2026 on the ElevenLabs realtime Speech to Text WebSocket as scribe_v3_live, in ElevenAgents, and as the default in the ElevenLabs app. It adds realtime speaker diarization for the first time, turn prediction that uses conversation context as well as silence (opt-in; the default commit strategy stays manual), a multilingual mode for mid-sentence language switches, and a keyterm library for vocabularies too large for a list. Upgrading from Scribe v2 Realtime is a model ID change; Scribe v2 Realtime remains available. ElevenLabs also announced a medical-conversation version for ElevenAgents.

## Other ElevenLabs Models

- [Scribe v3](/models/scribe-v3) - Score: not computed
- [ElevenLabs Eleven v4](/models/elevenlabs-eleven-v4) - Score: not computed
- [ElevenLabs Eleven v4 Turbo](/models/elevenlabs-eleven-v4-turbo) - Score: not computed
- [ElevenLabs Eleven v3 Conversational](/models/elevenlabs-eleven-v3-conversational) - Score: not computed
- [ElevenLabs Eleven v3](/models/elevenlabs-eleven-v3) - Score: not computed
- [Eleven Realtime v1 Max](/models/eleven-realtime-v1-max) - Score: not computed
- [Eleven Realtime v1 Mini](/models/eleven-realtime-v1-mini) - Score: not computed
- [Scribe v2.2 Realtime](/models/scribe-v2-2-realtime) - Score: not computed