ELEVENLABSSPEECH-TO-TEXT38,964 SAMPLES / 30 DAYSLAST RUN NOV 17, 2026, 03:00 UTC

Scribe v3 Live speech-to-text benchmarks

Scribe v3 Live, hosted by ElevenLabs (Eleven Labs, 11Labs), measures mean 98 ms time to final segment (10th of 30) and 2.0% word error rate (3rd of 31) among STT systems. Results cover the last 30 days. Last measured .

Coval has active speech-to-text measurements for Scribe v3 Live, created by ElevenLabs and served through ElevenLabs's own API.

Overview

Scribe v3 Live is tested every day on fixed public audio — clean, accented, noisy, reverberant, far-field, clipped and phone-codec speech — for transcription accuracy and streaming latency.

Technical specifications

Made by
ElevenLabs
Hosted by
ElevenLabs
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Keyterm biasing, Multilingual, VAD

How Scribe v3 Live ranks

Full STT dashboard
Time to Final Segment — every measured STT modelMilliseconds · lower is better · 30-day averageEvery STT model ranked on Time to Final Segment, with Scribe v3 Live highlighted.
  1. #1Wire-129 ms
  2. #2Qwen3 ASR 1.7b35 ms
  3. #3STT RT v551 ms
  4. #5STT Rt v659 ms
  5. #6STT 170 ms
  6. #7Nova 395 ms
  7. #8Nova 297 ms
  8. #9Rumble-298 ms
Show all 30 models
  1. #14Flux112 ms
  2. #15Scribe112 ms
  3. #17Whisper Large v3125 ms
  4. #18Ink 2141 ms
  5. #19Grok STT163 ms
  6. #20Pulse169 ms
  7. #21resonant-1210 ms
  8. #22Linden 1229 ms
  9. #24Default284 ms
  10. #25Daytona STT348 ms
  11. #27Solaria 1390 ms
  12. #29Chirp 3842 ms
  13. #30Chirp 2909 ms
median of all models · 114 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1Saidwire En 2026 10 30Saidwire29 ms768 ms9,441
2Qwen3 ASR 1.7bBaseten35 ms943 ms13,679
3STT RT v5Soniox51 ms1529 ms27,358
4Qwen3 ASR FastNari Labs53 ms1757 ms27,361
5STT Rt v6Soniox59 ms1893 ms12,188
6STT 1Inworld AI70 ms994 ms27,333
7Nova 3Deepgram95 ms1435 ms24,081
8Nova 2Deepgram97 ms1432 ms24,046
9Realtime GlobalReson898 ms1658 ms27,361
10Scribe v3 LiveElevenLabs98 ms1254 ms5,781
11Universal 3.5 ProAssemblyAI104 ms1013 ms27,361
12Universal 3.6 ProAssemblyAI110 ms1073 ms27,360
13Flux MultilingualDeepgram110 ms1173 ms24,079
14FluxDeepgram112 ms1095 ms24,085
15ScribeZoom112 ms—6,654
16Scribe v2 RealtimeElevenLabs116 ms2126 ms27,356
17Whisper Large v3Baseten125 ms920 ms13,221
18Ink 2Cartesia141 ms2030 ms18,412
19Grok STTSpaceXAI163 ms—27,343
20PulseSmallest169 ms1953 ms27,358
21resonant-1Reson8210 ms—27,353
22Linden 1Speechmatics229 ms1503 ms27,338
23Scribe Pro v2.1Zoom241 ms—6,823
24DefaultGradium284 ms2074 ms27,354
25Daytona STTGuava348 ms5609 ms13,376
26Gemini 3.5 Transcribe LiveGemini355 ms1520 ms27,282
27Solaria 1Gladia390 ms1534 ms27,318
28GPT Realtime WhisperOpenAI566 ms1813 ms26,878
29Chirp 3Google842 ms6035 ms27,311
30Chirp 2Google909 ms6101 ms27,359
—Universal StreamingAssemblyAI—1511 ms27,330
Under-sampled models are excluded; tied models share a place. Dotted WER values split into substitutions, deletions and insertions on hover or tap.

Highest relative placement: 3rd of 31 on Word Error Rate.

Latency vs accuracy

Where the errors come from

WER compositionScribe v3 Live's Word Error Rate split by error type · 30-day averageScribe v3 Live's WER split into substitutions, deletions and insertions.
  • Scribe v3 Live2.0%
SubstitutionsDeletionsInsertions

Averages and tail latency

Averages hide slow outliers — these are the distributions behind each figure.

Scribe v3 Live latency distributionMilliseconds · shared axis across metrics · last 30 daysScribe v3 Live's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to Final Segmentp50 90 ms · Max 14694 ms
  • Time to First Tokenp50 1285 ms · Max 4834 ms
band p25–p75 · tick p50 · whisker to max with p90 and p95 stops
Average and percentile values per metric, with the number of samples behind each row.
MetricAveragep25p50p75p90p95MaxSamples
Time to Final Segment98 ms77 ms90 ms106 ms123 ms137 ms14694 ms5,781
Word Error Rate2.0%0.0%0.0%3.1%8.3%11.8%50.0%5,781
Time to First Token1254 ms968 ms1285 ms1582 ms1964 ms2269 ms4834 ms5,781

Last 30 days

Daily medians from the same measurement runs · gaps are days without qualifying runs.

Time to Final Segment — daily p50Line p50 · band p25–p75 · UTC daysScribe v3 Live's daily median Time to Final Segment over the last 30 days.
Scribe v3 Live · Nov 10: 97 msScribe v3 Live · Nov 11: 85 msScribe v3 Live · Nov 12: 102 msScribe v3 Live · Nov 13: 96 msScribe v3 Live · Nov 14: 79 msScribe v3 Live · Nov 15: 87 msScribe v3 Live · Nov 16: 91 msScribe v3 Live · Nov 17: 77 ms
Word Error Rate — dailyDaily corpus-level · UTC daysScribe v3 Live's daily Word Error Rate over the last 30 days.
Scribe v3 Live · Nov 10: 3.8%Scribe v3 Live · Nov 11: 2.1%Scribe v3 Live · Nov 12: 2.0%Scribe v3 Live · Nov 13: 2.2%Scribe v3 Live · Nov 14: 1.9%Scribe v3 Live · Nov 15: 2.2%Scribe v3 Live · Nov 16: 2.0%Scribe v3 Live · Nov 17: 1.9%

Time to Final Segment by dataset

Scribe v3 Live by test conditionMilliseconds · lower is better · best condition firstScribe v3 Live's Time to Final Segment on each benchmark dataset.
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
ProductionPipeCat95 ms3,049
AccentsWildASR95 ms303
CleanWildASR108 ms1,216
ClippingWildASR101 ms304
Far-fieldWildASR99 ms302
Noise gapsWildASR95 ms304
Phone codecWildASR94 ms303

Strongest condition: WildASR phone codec at 94 ms · weakest: WildASR clean at 108 ms.

How fast is Scribe v3 Live?

On ElevenLabs, Scribe v3 Live measures mean 98 ms time to final segment (10th of 30) and mean 1254 ms time to first token (9th of 27). Last measured 2026-11-17.

How accurate is Scribe v3 Live?

On ElevenLabs, Scribe v3 Live measures 2.0% word error rate (3rd of 31). Last measured 2026-11-17.

Who hosts Scribe v3 Live?

Scribe v3 Live is created by ElevenLabs and served by ElevenLabs. Coval measures each hosted endpoint separately.

Results are re-measured daily using fixed datasets and reported over a rolling 30-day window. Same datasets, prompts and metric definitions for every model, measured by Coval’s open-source runner. Full methodology on the overview.

Evaluate your own voice agent

Use Coval to test your production configuration, prompts and calls—not only the public benchmark endpoints.

Book a Demo