Voice AI BenchmarksBy

ALIBABASPEECH-TO-TEXT136,855 SAMPLES / 30 DAYSLAST RUN NOV 2, 2026, 10:00 UTC

Qwen3 ASR Fast speech-to-text benchmarks

Qwen3 ASR Fast, hosted by Nari Labs, measures mean 50 ms time to final segment (3rd of 28) and 2.6% word error rate (3rd of 30) among STT systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Coval has active speech-to-text measurements for Qwen3 ASR Fast, created by Alibaba and served on Nari Labs.

Time to Final Segment#3 / 28
50ms
Word Error Rate#3 / 30
2.6%
Time to First Token#18 / 28
1755ms

Overview

Qwen3 ASR Fast 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
Alibaba
Hosted by
Nari Labs
Source
Shared inference
Licensing
Open-weight
Deployment
Cloud
Region
US
Features
Multilingual

How Qwen3 ASR Fast 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 Qwen3 ASR Fast highlighted.
  1. #1Qwen3 ASR 1.7b35 ms
  2. #2STT RT v544 ms
  3. #3Qwen3 ASR Fast50 ms
  4. #4STT 170 ms
  5. #5Nova 385 ms
  6. #6Nova 287 ms
  7. #7Parakeet TDT 0.6B v388 ms
  8. #8Flux Multilingual100 ms
  9. #9Flux101 ms
  10. #10Universal 3.6 Pro114 ms
  11. #11Scribe v2 Realtime118 ms
  12. #12Universal 3.5 Pro126 ms
Show all 28 models
  1. #13Whisper Large v3via Baseten126 ms
  2. #14Ink 2146 ms
  3. #15Realtime Global159 ms
  4. #16Whisper Large v3via Together AI175 ms
  5. #17Grok STT180 ms
  6. #18Pulse203 ms
  7. #19resonant-1243 ms
  8. #20Linden 1254 ms
  9. #21Default261 ms
  10. #22Gemini 3.5 Transcribe Live383 ms
  11. #23Daytona STT384 ms
  12. #24Solaria 1414 ms
  13. #25Voxtral Mini Transcribe Realtime 2602436 ms
  14. #26GPT Realtime Whisper563 ms
  15. #27Chirp 3807 ms
  16. #28Chirp 2896 ms
median of all models · 152 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1Qwen3 ASR 1.7bBaseten35 ms936 ms13,678
2STT RT v5Soniox44 ms1526 ms27,365
3Qwen3 ASR FastNari Labs50 ms1755 ms27,371
4STT 1Inworld AI70 ms1026 ms27,334
5Nova 3Deepgram85 ms1444 ms27,364
6Nova 2Deepgram87 ms1427 ms27,311
7Parakeet TDT 0.6B v3Together AI88 ms1198 ms15,388
8Flux MultilingualDeepgram100 ms1174 ms27,365
9FluxDeepgram101 ms1097 ms27,371
10Universal 3.6 ProAssemblyAI114 ms1080 ms27,335
11Scribe v2 RealtimeElevenLabs118 ms2151 ms27,367
12Universal 3.5 ProAssemblyAI126 ms1019 ms27,373
13Whisper Large v3Baseten126 ms914 ms13,674
14Ink 2Cartesia146 ms2034 ms27,369
15Realtime GlobalReson8159 ms2058 ms22,526
16Whisper Large v3Together AI175 ms1210 ms15,456
17Grok STTSpaceXAI180 ms—27,332
18PulseSmallest203 ms1988 ms27,366
19resonant-1Reson8243 ms—27,372
20Linden 1Speechmatics254 ms1528 ms27,368
21DefaultGradium261 ms2005 ms27,364
22Gemini 3.5 Transcribe LiveGemini383 ms1542 ms27,270
23Daytona STTGuava384 ms5651 ms13,133
24Solaria 1Gladia414 ms1608 ms27,306
25Voxtral Mini Transcribe Realtime 2602Mistral436 ms1868 ms20,723
26GPT Realtime WhisperOpenAI563 ms1822 ms26,862
27Chirp 3Google807 ms6008 ms27,330
28Chirp 2Google896 ms6097 ms27,368
—Nemotron 3.5 ASR StreamingTogether AI—1558 ms15,365
—Universal StreamingAssemblyAI—1516 ms27,339
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 30 on Word Error Rate.

Latency vs accuracy

TTFS vs WEREach point is one measured model · Qwen3 ASR Fast highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with Qwen3 ASR Fast highlighted.
Qwen3 ASR 1.7b — 35 ms, 2.9%STT RT v5 — 44 ms, 4.8%STT 1 — 70 ms, 4.4%Nova 3 — 85 ms, 5.2%Nova 2 — 87 ms, 6.6%Parakeet TDT 0.6B v3 — 88 ms, 10.9%Flux Multilingual — 100 ms, 7.4%Flux — 101 ms, 5.7%Universal 3.6 Pro — 114 ms, 2.2%Scribe v2 Realtime — 118 ms, 4.2%Universal 3.5 Pro — 126 ms, 2.4%Whisper Large v3 via Baseten — 126 ms, 4.5%Ink 2 — 146 ms, 3.4%Realtime Global — 159 ms, 2.6%Whisper Large v3 via Together AI — 175 ms, 7.2%Grok STT — 180 ms, 4.1%Pulse — 203 ms, 4.6%resonant-1 — 243 ms, 2.6%Linden 1 — 254 ms, 3.8%Default — 261 ms, 8.4%Gemini 3.5 Transcribe Live — 383 ms, 2.9%Daytona STT — 384 ms, 4.8%Solaria 1 — 414 ms, 6.2%Voxtral Mini Transcribe Realtime 2602 — 436 ms, 5.3%GPT Realtime Whisper — 563 ms, 4.0%Chirp 3 — 807 ms, 3.2%Chirp 2 — 896 ms, 3.9%Qwen3 ASR Fast — 50 ms, 2.6%

Where the errors come from

WER compositionQwen3 ASR Fast's Word Error Rate split by error type · 30-day averageQwen3 ASR Fast's WER split into substitutions, deletions and insertions.
  • Qwen3 ASR Fast2.6%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Qwen3 ASR Fast latency distributionMilliseconds · shared axis across metrics · last 30 daysQwen3 ASR Fast's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to Final Segmentp50 46 ms · Max 30057 ms
  • Time to First Tokenp50 1750 ms · Max 8639 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 Segment50 ms40 ms46 ms51 ms61 ms70 ms30057 ms27,371
Word Error Rate2.6%0.0%0.0%4.3%10.0%14.3%57.1%27,371
Time to First Token1755 ms1747 ms1750 ms1755 ms1763 ms1775 ms8639 ms27,371

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 daysQwen3 ASR Fast's daily median Time to Final Segment over the last 30 days.
Qwen3 ASR Fast · Oct 3: 47 msQwen3 ASR Fast · Oct 4: 43 msQwen3 ASR Fast · Oct 5: 44 msQwen3 ASR Fast · Oct 6: 42 msQwen3 ASR Fast · Oct 7: 44 msQwen3 ASR Fast · Oct 8: 46 msQwen3 ASR Fast · Oct 9: 46 msQwen3 ASR Fast · Oct 10: 43 msQwen3 ASR Fast · Oct 11: 42 msQwen3 ASR Fast · Oct 12: 46 msQwen3 ASR Fast · Oct 13: 45 msQwen3 ASR Fast · Oct 14: 47 msQwen3 ASR Fast · Oct 15: 46 msQwen3 ASR Fast · Oct 16: 47 msQwen3 ASR Fast · Oct 17: 43 msQwen3 ASR Fast · Oct 18: 44 msQwen3 ASR Fast · Oct 19: 48 msQwen3 ASR Fast · Oct 20: 44 msQwen3 ASR Fast · Oct 21: 43 msQwen3 ASR Fast · Oct 22: 48 msQwen3 ASR Fast · Oct 23: 49 msQwen3 ASR Fast · Oct 24: 45 msQwen3 ASR Fast · Oct 25: 43 msQwen3 ASR Fast · Oct 26: 49 msQwen3 ASR Fast · Oct 27: 43 msQwen3 ASR Fast · Oct 28: 46 msQwen3 ASR Fast · Oct 29: 51 msQwen3 ASR Fast · Oct 30: 50 msQwen3 ASR Fast · Oct 31: 50 msQwen3 ASR Fast · Nov 1: 47 msQwen3 ASR Fast · Nov 2: 53 ms
Word Error Rate — daily averageDaily average · UTC daysQwen3 ASR Fast's daily Word Error Rate over the last 30 days.
Qwen3 ASR Fast · Oct 3: 3.0%Qwen3 ASR Fast · Oct 4: 3.9%Qwen3 ASR Fast · Oct 5: 3.1%Qwen3 ASR Fast · Oct 6: 4.2%Qwen3 ASR Fast · Oct 7: 3.8%Qwen3 ASR Fast · Oct 8: 3.0%Qwen3 ASR Fast · Oct 9: 3.5%Qwen3 ASR Fast · Oct 10: 3.7%Qwen3 ASR Fast · Oct 11: 2.6%Qwen3 ASR Fast · Oct 12: 3.4%Qwen3 ASR Fast · Oct 13: 3.4%Qwen3 ASR Fast · Oct 14: 3.2%Qwen3 ASR Fast · Oct 15: 3.9%Qwen3 ASR Fast · Oct 16: 4.3%Qwen3 ASR Fast · Oct 17: 3.7%Qwen3 ASR Fast · Oct 18: 3.6%Qwen3 ASR Fast · Oct 19: 3.2%Qwen3 ASR Fast · Oct 20: 4.8%Qwen3 ASR Fast · Oct 21: 3.7%Qwen3 ASR Fast · Oct 22: 3.8%Qwen3 ASR Fast · Oct 23: 3.2%Qwen3 ASR Fast · Oct 24: 3.8%Qwen3 ASR Fast · Oct 25: 3.4%Qwen3 ASR Fast · Oct 26: 3.2%Qwen3 ASR Fast · Oct 27: 4.3%Qwen3 ASR Fast · Oct 28: 3.5%Qwen3 ASR Fast · Oct 29: 2.9%Qwen3 ASR Fast · Oct 30: 3.3%Qwen3 ASR Fast · Oct 31: 3.6%Qwen3 ASR Fast · Nov 1: 3.9%Qwen3 ASR Fast · Nov 2: 4.6%

Time to Final Segment by dataset

Qwen3 ASR Fast by test conditionMilliseconds · lower is better · best condition firstQwen3 ASR Fast's Time to Final Segment on each benchmark dataset.
  1. WildASR accents47 ms
  2. WildASR noise gaps47 ms
  3. WildASR clipping48 ms
  4. WildASR far-field49 ms
  5. WildASR phone codec49 ms
  6. PipeCat (production)50 ms
  7. WildASR clean54 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
ProductionPipeCat50 ms14,409
AccentsWildASR47 ms1,441
CleanWildASR54 ms5,759
ClippingWildASR48 ms1,442
Far-fieldWildASR49 ms1,440
Noise gapsWildASR47 ms1,440
Phone codecWildASR49 ms1,440

Strongest condition: WildASR accents at 47 ms · weakest: WildASR clean at 54 ms.

How fast is Qwen3 ASR Fast?

On Nari Labs, Qwen3 ASR Fast measures mean 50 ms time to final segment (3rd of 28) and mean 1755 ms time to first token (18th of 28). Last measured 2026-11-02.

How accurate is Qwen3 ASR Fast?

On Nari Labs, Qwen3 ASR Fast measures 2.6% word error rate (3rd of 30). Last measured 2026-11-02.

Who hosts Qwen3 ASR Fast?

Qwen3 ASR Fast is created by Alibaba and served by Nari Labs. 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