Voice AI BenchmarksBy

INWORLD AISPEECH-TO-TEXT123,740 SAMPLES / 30 DAYSLAST RUN OCT 14, 2026, 15:30 UTC

official resource

STT 1 speech-to-text benchmarks

STT 1, hosted by Inworld AI, measures mean 67 ms time to final segment (4th of 29) and 7.1% word error rate (26th of 31) among STT systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Inworld STT 1 is Inworld's streaming transcription model for interactive systems.

Time to Final Segment#4 / 29
67ms
Word Error Rate#26 / 31
7.1%
Time to First Token#7 / 27
1206ms

Overview

The tested endpoint is multilingual and includes voice activity plus keyterm biasing.

STT 1 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
Inworld AI
Hosted by
Inworld AI
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Keyterm biasing, Multilingual, VAD

How STT 1 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 STT 1 highlighted.
  1. #1Qwen3 ASR 1.7b39 ms
  2. #2STT RT v545 ms
  3. #3Qwen3 ASR Fast47 ms
  4. #4STT 167 ms
  5. #5Nova 385 ms
  6. #6Nova 287 ms
  7. #7Parakeet TDT 0.6B v397 ms
  8. #8Flux Multilingual98 ms
  9. #9Flux99 ms
  10. #10Scribe v2 Realtime121 ms
  11. #11Whisper Large v3via Baseten131 ms
  12. #12Universal 3.6 Pro135 ms
Show all 29 models
  1. #13Ink 2149 ms
  2. #14Universal 3.5 Pro170 ms
  3. #15Whisper Large v3via Together AI198 ms
  4. #16Pulse205 ms
  5. #17Grok STT223 ms
  6. #18resonant-1252 ms
  7. #19Default261 ms
  8. #20Linden 1274 ms
  9. #21Gemini 3.5 Transcribe Live317 ms
  10. #22Daytona STT399 ms
  11. #23Voxtral Mini Transcribe Realtime 2602415 ms
  12. #24GPT Realtime Whisper599 ms
  13. #25Solaria 1621 ms
  14. #26GPT-4o mini Transcribe709 ms
  15. #27Chirp 3768 ms
  16. #28GPT-4o Transcribe783 ms
  17. #29Chirp 2909 ms
median of all models · 198 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1Qwen3 ASR 1.7bBaseten39 ms946 ms13,669
2STT RT v5Soniox45 ms1518 ms23,533
3Qwen3 ASR FastNari Labs47 ms1752 ms26,652
4STT 1Inworld AI67 ms1206 ms24,748
5Nova 3Deepgram85 ms1432 ms26,002
6Nova 2Deepgram87 ms1417 ms25,960
7Parakeet TDT 0.6B v3Together AI97 ms1228 ms26,461
8Flux MultilingualDeepgram98 ms1165 ms26,007
9FluxDeepgram99 ms1091 ms26,006
10Scribe v2 RealtimeElevenLabs121 ms2157 ms26,640
11Whisper Large v3Baseten131 ms924 ms13,663
12Universal 3.6 ProAssemblyAI135 ms1071 ms24,546
13Ink 2Cartesia149 ms1969 ms26,653
14Universal 3.5 ProAssemblyAI170 ms1026 ms26,654
15Whisper Large v3Together AI198 ms1258 ms26,448
16PulseSmallest205 ms1976 ms26,654
17Grok STTSpaceXAI223 ms—26,577
18resonant-1Reson8252 ms—26,653
19DefaultGradium261 ms1987 ms26,632
20Linden 1Speechmatics274 ms1537 ms26,595
21Gemini 3.5 Transcribe LiveGemini317 ms1603 ms26,566
22Daytona STTGuava399 ms5695 ms5,225
23Voxtral Mini Transcribe Realtime 2602Mistral415 ms1852 ms26,071
24GPT Realtime WhisperOpenAI599 ms1824 ms26,622
25Solaria 1Gladia621 ms1769 ms26,309
26GPT-4o mini TranscribeOpenAI709 ms—26,621
27Chirp 3Google768 ms5956 ms26,638
28GPT-4o TranscribeOpenAI783 ms—26,625
29Chirp 2Google909 ms6099 ms26,650
—Nemotron 3.5 ASR StreamingTogether AI—1544 ms26,549
—Universal StreamingAssemblyAI—1509 ms26,626
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

TTFS vs WEREach point is one measured model · STT 1 highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with STT 1 highlighted.
Qwen3 ASR 1.7b — 39 ms, 2.9%STT RT v5 — 45 ms, 4.8%Qwen3 ASR Fast — 47 ms, 2.6%Nova 3 — 85 ms, 5.3%Nova 2 — 87 ms, 6.7%Parakeet TDT 0.6B v3 — 97 ms, 11.1%Flux Multilingual — 98 ms, 7.5%Flux — 99 ms, 5.8%Scribe v2 Realtime — 121 ms, 4.4%Whisper Large v3 via Baseten — 131 ms, 4.5%Universal 3.6 Pro — 135 ms, 2.1%Ink 2 — 149 ms, 3.6%Universal 3.5 Pro — 170 ms, 2.4%Whisper Large v3 via Together AI — 198 ms, 7.4%Pulse — 205 ms, 4.6%Grok STT — 223 ms, 4.1%resonant-1 — 252 ms, 2.6%Default — 261 ms, 8.8%Linden 1 — 274 ms, 3.9%Gemini 3.5 Transcribe Live — 317 ms, 3.0%Daytona STT — 399 ms, 4.9%Voxtral Mini Transcribe Realtime 2602 — 415 ms, 5.4%GPT Realtime Whisper — 599 ms, 4.0%Solaria 1 — 621 ms, 6.0%GPT-4o mini Transcribe — 709 ms, 3.9%Chirp 3 — 768 ms, 3.2%GPT-4o Transcribe — 783 ms, 3.8%Chirp 2 — 909 ms, 3.9%STT 1 — 67 ms, 7.1%

Where the errors come from

WER compositionSTT 1's Word Error Rate split by error type · 30-day averageSTT 1's WER split into substitutions, deletions and insertions.
  • STT 17.1%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

STT 1 latency distributionMilliseconds · shared axis across metrics · last 30 daysSTT 1's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to Final Segmentp50 64 ms · Max 4699 ms
  • Time to First Tokenp50 1046 ms · Max 13056 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 Segment67 ms57 ms64 ms72 ms85 ms101 ms4699 ms24,748
Word Error Rate7.1%0.0%0.0%7.1%15.8%28.6%646.2%24,748
Time to First Token1206 ms1035 ms1046 ms1337 ms1636 ms1737 ms13056 ms24,748

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 daysSTT 1's daily median Time to Final Segment over the last 30 days.
STT 1 · Sep 14: 67 msSTT 1 · Sep 15: 65 msSTT 1 · Sep 16: 64 msSTT 1 · Sep 17: 79 msSTT 1 · Sep 18: 69 msSTT 1 · Sep 19: 60 msSTT 1 · Sep 20: 64 msSTT 1 · Sep 21: 60 msSTT 1 · Sep 22: 61 msSTT 1 · Sep 23: 64 msSTT 1 · Sep 24: 61 msSTT 1 · Sep 25: 67 msSTT 1 · Sep 26: 60 msSTT 1 · Sep 27: 61 msSTT 1 · Sep 30: 71 msSTT 1 · Oct 1: 65 msSTT 1 · Oct 2: 67 msSTT 1 · Oct 3: 65 msSTT 1 · Oct 4: 69 msSTT 1 · Oct 5: 62 msSTT 1 · Oct 6: 65 msSTT 1 · Oct 7: 69 msSTT 1 · Oct 8: 64 msSTT 1 · Oct 9: 65 msSTT 1 · Oct 10: 63 msSTT 1 · Oct 11: 65 msSTT 1 · Oct 12: 65 msSTT 1 · Oct 13: 67 msSTT 1 · Oct 14: 66 ms
Word Error Rate — daily averageDaily average · UTC daysSTT 1's daily Word Error Rate over the last 30 days.
STT 1 · Sep 14: 3.1%STT 1 · Sep 15: 4.0%STT 1 · Sep 16: 28.4%STT 1 · Sep 17: 63.4%STT 1 · Sep 18: 54.0%STT 1 · Sep 19: 4.5%STT 1 · Sep 20: 4.4%STT 1 · Sep 21: 4.3%STT 1 · Sep 22: 4.4%STT 1 · Sep 23: 4.3%STT 1 · Sep 24: 8.3%STT 1 · Sep 25: 5.4%STT 1 · Sep 26: 7.9%STT 1 · Sep 27: 7.3%STT 1 · Sep 28: 3.6%STT 1 · Sep 30: 6.4%STT 1 · Oct 1: 3.4%STT 1 · Oct 2: 4.9%STT 1 · Oct 3: 4.5%STT 1 · Oct 4: 8.3%STT 1 · Oct 5: 4.8%STT 1 · Oct 6: 13.8%STT 1 · Oct 7: 4.4%STT 1 · Oct 8: 7.2%STT 1 · Oct 9: 8.6%STT 1 · Oct 10: 8.0%STT 1 · Oct 11: 3.1%STT 1 · Oct 12: 8.5%STT 1 · Oct 13: 4.9%STT 1 · Oct 14: 5.4%

Time to Final Segment by dataset

STT 1 by test conditionMilliseconds · lower is better · best condition firstSTT 1's Time to Final Segment on each benchmark dataset.
  1. WildASR accents58 ms
  2. PipeCat (production)65 ms
  3. WildASR phone codec65 ms
  4. WildASR clipping70 ms
  5. WildASR noise gaps70 ms
  6. WildASR clean70 ms
  7. WildASR far-field71 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
ProductionPipeCat65 ms13,014
AccentsWildASR58 ms1,305
CleanWildASR70 ms5,215
ClippingWildASR70 ms1,304
Far-fieldWildASR71 ms1,303
Noise gapsWildASR70 ms1,304
Phone codecWildASR65 ms1,303

Strongest condition: WildASR accents at 58 ms · weakest: WildASR far-field at 71 ms.

How fast is STT 1?

On Inworld AI, STT 1 measures mean 67 ms time to final segment (4th of 29) and mean 1206 ms time to first token (7th of 27). Last measured 2026-10-14.

How accurate is STT 1?

On Inworld AI, STT 1 measures 7.1% word error rate (26th of 31). Last measured 2026-10-14.

Who hosts STT 1?

STT 1 is created by Inworld AI and served by Inworld AI. Coval measures each hosted endpoint separately.

Limits of this comparison

Coval measures it independently from Inworld's TTS 2 family so recognition and synthesis remain attributable to the correct stage.

  • Coval does not evaluate game-character behavior or other application features in Inworld's broader platform.

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