Voice AI BenchmarksBy

ELEVENLABSTEXT-TO-SPEECH36,652 SAMPLES / 30 DAYSLAST RUN SEP 15, 2026, 09:00 UTC

official resource

Flash v2.5 text-to-speech benchmarks

Flash v2.5, hosted by ElevenLabs (Eleven Labs, 11Labs), measures mean 194 ms time to first audio (7th of 28) and 6.7% word error rate (27th of 28) among TTS systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Eleven Flash v2.5 is ElevenLabs' latency-oriented multilingual synthesis model.

Time to First Audio#7 / 28
194ms
TTFA Network Roundtrip#11 / 28
154ms
TTFA Leading Silence#11 / 28
40ms
Word Error Rate#27 / 28
6.7%

Overview

Flash v2.5 is built for real-time use with voice cloning; network and leading-silence timing components are reported separately where available.

Flash v2.5 is tested every day on a fixed set of text prompts, measuring how quickly audible speech starts and how intelligible the result is.

Technical specifications

Made by
ElevenLabs
Hosted by
ElevenLabs
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Multilingual, Voice cloning

How Flash v2.5 ranks

Full TTS dashboard
Time to First Audio — every measured TTS modelMilliseconds · lower is better · 30-day averageEvery TTS model ranked on Time to First Audio, with Flash v2.5 highlighted.
  1. #1vui66 ms
  2. #2Qwen3 TTS Fast71 ms
  3. #3TTS Flash 291 ms
  4. #4Qwen3 TTS 1.7b106 ms
  5. #5Palabra TTS v1115 ms
  6. #6TTS 2181 ms
  7. #7Flash v2.5194 ms
  8. #8Default235 ms
  9. #9TTS Rt v2245 ms
  10. #10TTS RT v1245 ms
  11. #11Mist v3258 ms
  12. #12Sonic 3.5276 ms
Show all 28 models
  1. #13Phantom Z 3.4 conversational286 ms
  2. #14Aura 2308 ms
  3. #15Coda312 ms
  4. #16S2.1 Pro335 ms
  5. #17Eleven v3 Conversational348 ms
  6. #18Lightning v3.1 Pro362 ms
  7. #19S1379 ms
  8. #20Grok TTS397 ms
  9. #21Sonic 3.6423 ms
  10. #22Simba 3.2451 ms
  11. #23Simba 3.0473 ms
  12. #24Chirp 3 HD535 ms
  13. #25Falcon 2545 ms
  14. #26Qwen3 TTS Flash Realtime753 ms
  15. #27S2.1 Pro Free963 ms
  16. #28GPT-4o mini TTS1014 ms
median of all models · 310 ms
TTS models over the last 30 days, ranked on Time to First Audio.
#ModelHostTTFAWERSamples
1vuiFluxions66 ms9,500
2Qwen3 TTS FastNari71 ms2,240
3TTS Flash 2Inworld AI91 ms11,451
4Qwen3 TTS 1.7bBaseten106 ms420
5Palabra TTS v1Palabra115 ms11,435
6TTS 2Inworld AI181 ms11,451
7Flash v2.5ElevenLabs194 ms9,168
8DefaultGradium235 ms9,177
9TTS Rt v2Soniox245 ms11,232
10TTS RT v1Soniox245 ms11,234
11Mist v3Rime258 ms11,437
12Sonic 3.5Cartesia276 ms11,450
13Phantom Z 3.4 conversationalDeepdub286 ms11,446
14Aura 2Deepgram308 ms11,436
15CodaRime312 ms11,442
16S2.1 ProFish Audio335 ms11,450
17Eleven v3 ConversationalElevenLabs348 ms11,439
18Lightning v3.1 ProSmallest362 ms11,451
19S1Fish Audio379 ms11,443
20Grok TTSxAI397 ms11,436
21Sonic 3.6Cartesia423 ms8,260
22Simba 3.2Speechify451 ms11,443
23Simba 3.0Speechify473 ms11,446
24Chirp 3 HDGoogle535 ms11,450
25Falcon 2Murf545 ms11,447
26Qwen3 TTS Flash RealtimeAlibaba753 ms11,276
27S2.1 Pro FreeFish Audio963 ms11,416
28GPT-4o mini TTSOpenAI1014 ms11,420
Under-sampled models are excluded; tied models share a place. Dotted WER values split into substitutions, deletions and insertions on hover or tap.

Latency vs accuracy

TTFA vs WEREach point is one measured model · Flash v2.5 highlighted · 30-day averagesTime to First Audio against Word Error Rate for every measured TTS model, with Flash v2.5 highlighted.
vui — 66 ms, 5.5%Qwen3 TTS Fast — 71 ms, 4.2%TTS Flash 2 — 91 ms, 5.4%Qwen3 TTS 1.7b — 106 ms, 5.5%Palabra TTS v1 — 115 ms, 5.8%TTS 2 — 181 ms, 4.6%Default — 235 ms, 5.3%TTS Rt v2 — 245 ms, 4.2%TTS RT v1 — 245 ms, 3.9%Mist v3 — 258 ms, 5.6%Sonic 3.5 — 276 ms, 5.9%Phantom Z 3.4 conversational — 286 ms, 6.0%Aura 2 — 308 ms, 5.2%Coda — 312 ms, 5.0%S2.1 Pro — 335 ms, 4.8%Eleven v3 Conversational — 348 ms, 4.4%Lightning v3.1 Pro — 362 ms, 4.4%S1 — 379 ms, 4.9%Grok TTS — 397 ms, 4.9%Sonic 3.6 — 423 ms, 5.3%Simba 3.2 — 451 ms, 4.4%Simba 3.0 — 473 ms, 5.2%Chirp 3 HD — 535 ms, 5.2%Falcon 2 — 545 ms, 5.2%Qwen3 TTS Flash Realtime — 753 ms, 8.9%S2.1 Pro Free — 963 ms, 4.7%GPT-4o mini TTS — 1014 ms, 4.8%Flash v2.5 — 194 ms, 6.7%

Where the errors come from

WER compositionFlash v2.5's Word Error Rate split by error type · 30-day averageFlash v2.5's WER split into substitutions, deletions and insertions.
  • Flash v2.56.7%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Flash v2.5 latency distributionMilliseconds · shared axis across metrics · last 30 daysFlash v2.5's latency percentiles per metric: p25–p75 band, p50 tick, whisker to p99.
  • Time to First Audiop50 186 ms · p99 397 ms
  • TTFA Network Roundtripp50 143 ms · p99 362 ms
  • TTFA Leading Silencep50 39 ms · p99 86 ms
band p25–p75 · tick p50 · whisker to p99 with p90 and p95 stops
Average and percentile values per metric, with the number of samples behind each row.
MetricAveragep25p50p75p90p95p99Samples
Time to First Audio194 ms173 ms186 ms200 ms217 ms231 ms397 ms9,168
TTFA Network Roundtrip154 ms138 ms143 ms152 ms167 ms181 ms362 ms9,168
TTFA Leading Silence40 ms31 ms39 ms50 ms60 ms68 ms86 ms9,168
Word Error Rate6.7%0.0%0.0%10.5%21.4%31.3%40.0%9,148

Last 30 days

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

Time to First Audio — daily p50Line p50 · band p25–p75 · UTC daysFlash v2.5's daily median Time to First Audio over the last 30 days.
Flash v2.5 · Sep 1: 182 msFlash v2.5 · Sep 2: 187 msFlash v2.5 · Sep 3: 187 msFlash v2.5 · Sep 4: 185 msFlash v2.5 · Sep 5: 182 msFlash v2.5 · Sep 6: 183 msFlash v2.5 · Sep 7: 185 msFlash v2.5 · Sep 8: 187 msFlash v2.5 · Sep 9: 184 msFlash v2.5 · Sep 10: 184 msFlash v2.5 · Sep 11: 186 msFlash v2.5 · Sep 12: 184 msFlash v2.5 · Sep 13: 186 msFlash v2.5 · Sep 14: 186 msFlash v2.5 · Sep 15: 183 ms
Word Error Rate — daily averageDaily average · UTC daysFlash v2.5's daily Word Error Rate over the last 30 days.
Flash v2.5 · Sep 1: 4.9%Flash v2.5 · Sep 2: 6.2%Flash v2.5 · Sep 3: 6.1%Flash v2.5 · Sep 4: 6.9%Flash v2.5 · Sep 5: 7.9%Flash v2.5 · Sep 6: 8.5%Flash v2.5 · Sep 7: 6.3%Flash v2.5 · Sep 8: 7.1%Flash v2.5 · Sep 9: 6.7%Flash v2.5 · Sep 10: 8.0%Flash v2.5 · Sep 11: 7.4%Flash v2.5 · Sep 12: 7.0%Flash v2.5 · Sep 13: 6.0%Flash v2.5 · Sep 14: 6.6%Flash v2.5 · Sep 15: 7.2%

Time to First Audio by dataset

Flash v2.5 by test conditionMilliseconds · lower is better · best condition firstFlash v2.5's Time to First Audio on each benchmark dataset.
  1. Text prompts194 ms
Time to First Audio per dataset, with the number of samples behind each figure.
DatasetTTFASamples
Text prompts194 ms9,168

Strongest condition: Text prompts at 194 ms · weakest: Text prompts at 194 ms.

How fast is Flash v2.5?

On ElevenLabs, Flash v2.5 measures mean 194 ms time to first audio (7th of 28). Last measured 2026-09-15.

How accurate is Flash v2.5?

On ElevenLabs, Flash v2.5 measures 6.7% word error rate (27th of 28). Last measured 2026-09-15.

Who hosts Flash v2.5?

Flash v2.5 is created by ElevenLabs and served by ElevenLabs. Coval measures each hosted endpoint separately.

Limits of this comparison

Coval's timing begins at the request and ends at audible speech, matching what a voice-agent caller experiences.

  • The public benchmark does not rank subjective naturalness, emotion fidelity or clone similarity.

Official sources

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