Voice AI BenchmarksBy

RESON8SPEECH-TO-TEXT112,820 SAMPLES / 30 DAYSLAST RUN NOV 2, 2026, 11:00 UTC

Realtime Global speech-to-text benchmarks

Realtime Global, hosted by Reson8, measures mean 159 ms time to final segment (15th of 28) and 2.6% word error rate (4th 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 Realtime Global, created by Reson8 and served through Reson8's own API.

Time to Final Segment#15 / 28
159ms
Word Error Rate#4 / 30
2.6%
Time to First Token#24 / 28
2056ms

Overview

Realtime Global 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
Reson8
Hosted by
Reson8
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Diarization, Keyterm biasing, Multilingual

How Realtime Global 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 Realtime Global 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 v389 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
  13. #13Whisper Large v3via Baseten126 ms
  14. #14Ink 2146 ms
  15. #15Realtime Global159 ms
Show all 28 models
  1. #16Whisper Large v3via Together AI174 ms
  2. #17Grok STT180 ms
  3. #18Pulse203 ms
  4. #19resonant-1243 ms
  5. #20Linden 1254 ms
  6. #21Default261 ms
  7. #22Gemini 3.5 Transcribe Live383 ms
  8. #23Daytona STT384 ms
  9. #24Solaria 1414 ms
  10. #25Voxtral Mini Transcribe Realtime 2602437 ms
  11. #26GPT Realtime Whisper563 ms
  12. #27Chirp 3807 ms
  13. #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 ms1525 ms27,364
3Qwen3 ASR FastNari Labs50 ms1755 ms27,371
4STT 1Inworld AI70 ms1026 ms27,334
5Nova 3Deepgram85 ms1444 ms27,364
6Nova 2Deepgram87 ms1426 ms27,311
7Parakeet TDT 0.6B v3Together AI89 ms1198 ms15,350
8Flux MultilingualDeepgram100 ms1174 ms27,365
9FluxDeepgram101 ms1097 ms27,371
10Universal 3.6 ProAssemblyAI114 ms1079 ms27,335
11Scribe v2 RealtimeElevenLabs118 ms2151 ms27,366
12Universal 3.5 ProAssemblyAI126 ms1019 ms27,373
13Whisper Large v3Baseten126 ms914 ms13,674
14Ink 2Cartesia146 ms2034 ms27,369
15Realtime GlobalReson8159 ms2056 ms22,564
16Whisper Large v3Together AI174 ms1210 ms15,418
17Grok STTSpaceXAI180 ms—27,332
18PulseSmallest203 ms1988 ms27,366
19resonant-1Reson8243 ms—27,372
20Linden 1Speechmatics254 ms1527 ms27,368
21DefaultGradium261 ms2005 ms27,364
22Gemini 3.5 Transcribe LiveGemini383 ms1541 ms27,270
23Daytona STTGuava384 ms5651 ms13,133
24Solaria 1Gladia414 ms1608 ms27,306
25Voxtral Mini Transcribe Realtime 2602Mistral437 ms1868 ms20,721
26GPT Realtime WhisperOpenAI563 ms1822 ms26,862
27Chirp 3Google807 ms6008 ms27,330
28Chirp 2Google896 ms6097 ms27,368
—Nemotron 3.5 ASR StreamingTogether AI—1558 ms15,327
—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: 4th of 30 on Word Error Rate.

Latency vs accuracy

TTFS vs WEREach point is one measured model · Realtime Global highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with Realtime Global highlighted.
Qwen3 ASR 1.7b — 35 ms, 2.9%STT RT v5 — 44 ms, 4.7%Qwen3 ASR Fast — 50 ms, 2.6%STT 1 — 70 ms, 4.4%Nova 3 — 85 ms, 5.2%Nova 2 — 87 ms, 6.6%Parakeet TDT 0.6B v3 — 89 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%Whisper Large v3 via Together AI — 174 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 — 437 ms, 5.3%GPT Realtime Whisper — 563 ms, 4.0%Chirp 3 — 807 ms, 3.1%Chirp 2 — 896 ms, 3.9%Realtime Global — 159 ms, 2.6%

Where the errors come from

WER compositionRealtime Global's Word Error Rate split by error type · 30-day averageRealtime Global's WER split into substitutions, deletions and insertions.
  • Realtime Global2.6%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Realtime Global latency distributionMilliseconds · shared axis across metrics · last 30 daysRealtime Global's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to Final Segmentp50 158 ms · Max 4966 ms
  • Time to First Tokenp50 2329 ms · Max 5665 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 Segment159 ms145 ms158 ms172 ms176 ms185 ms4966 ms22,564
Word Error Rate2.6%0.0%0.0%4.3%9.5%14.3%100.0%22,564
Time to First Token2056 ms1441 ms2329 ms2330 ms2335 ms3328 ms5665 ms22,564

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 daysRealtime Global's daily median Time to Final Segment over the last 30 days.
Realtime Global · Oct 8: 160 msRealtime Global · Oct 9: 165 msRealtime Global · Oct 10: 163 msRealtime Global · Oct 11: 164 msRealtime Global · Oct 12: 163 msRealtime Global · Oct 13: 163 msRealtime Global · Oct 14: 163 msRealtime Global · Oct 15: 173 msRealtime Global · Oct 16: 163 msRealtime Global · Oct 17: 158 msRealtime Global · Oct 18: 164 msRealtime Global · Oct 19: 157 msRealtime Global · Oct 20: 170 msRealtime Global · Oct 21: 163 msRealtime Global · Oct 22: 164 msRealtime Global · Oct 23: 164 msRealtime Global · Oct 24: 164 msRealtime Global · Oct 25: 164 msRealtime Global · Oct 26: 163 msRealtime Global · Oct 27: 164 msRealtime Global · Oct 28: 165 msRealtime Global · Oct 29: 171 msRealtime Global · Oct 30: 118 msRealtime Global · Oct 31: 115 msRealtime Global · Nov 1: 115 msRealtime Global · Nov 2: 113 ms
Word Error Rate — daily averageDaily average · UTC daysRealtime Global's daily Word Error Rate over the last 30 days.
Realtime Global · Oct 8: 3.6%Realtime Global · Oct 9: 3.3%Realtime Global · Oct 10: 3.1%Realtime Global · Oct 11: 2.9%Realtime Global · Oct 12: 3.6%Realtime Global · Oct 13: 3.6%Realtime Global · Oct 14: 3.8%Realtime Global · Oct 15: 3.8%Realtime Global · Oct 16: 4.2%Realtime Global · Oct 17: 3.6%Realtime Global · Oct 18: 3.1%Realtime Global · Oct 19: 4.2%Realtime Global · Oct 20: 4.7%Realtime Global · Oct 21: 3.8%Realtime Global · Oct 22: 4.2%Realtime Global · Oct 23: 3.2%Realtime Global · Oct 24: 4.1%Realtime Global · Oct 25: 3.4%Realtime Global · Oct 26: 3.4%Realtime Global · Oct 27: 4.6%Realtime Global · Oct 28: 3.7%Realtime Global · Oct 29: 3.6%Realtime Global · Oct 30: 3.2%Realtime Global · Oct 31: 2.7%Realtime Global · Nov 1: 3.2%Realtime Global · Nov 2: 3.1%

Time to Final Segment by dataset

Realtime Global by test conditionMilliseconds · lower is better · best condition firstRealtime Global's Time to Final Segment on each benchmark dataset.
  1. WildASR accents148 ms
  2. WildASR clean159 ms
  3. WildASR noise gaps159 ms
  4. PipeCat (production)159 ms
  5. WildASR clipping161 ms
  6. WildASR phone codec161 ms
  7. WildASR far-field161 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
ProductionPipeCat159 ms11,879
AccentsWildASR148 ms1,188
CleanWildASR159 ms4,748
ClippingWildASR161 ms1,188
Far-fieldWildASR161 ms1,187
Noise gapsWildASR159 ms1,187
Phone codecWildASR161 ms1,187

Strongest condition: WildASR accents at 148 ms · weakest: WildASR far-field at 161 ms.

How fast is Realtime Global?

On Reson8, Realtime Global measures mean 159 ms time to final segment (15th of 28) and mean 2056 ms time to first token (24th of 28). Last measured 2026-11-02.

How accurate is Realtime Global?

On Reson8, Realtime Global measures 2.6% word error rate (4th of 30). Last measured 2026-11-02.

Who hosts Realtime Global?

Realtime Global is created by Reson8 and served by Reson8. 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