Voice AI BenchmarksBy

OPENAISPEECH-TO-TEXT146,594 SAMPLES / 30 DAYSLAST RUN AUG 28, 2026, 21:30 UTC

official resource

GPT Realtime Whisper speech-to-text benchmarks

GPT Realtime Whisper is Whisper transcription running inside an OpenAI Realtime session, rather than the standalone Audio API.

Time to Final Segment#19 / 24
559ms
Word Error Rate#11 / 28
5.2%
Time to First Token#18 / 24
1823ms

Overview

The path runs within a persistent Realtime connection while still being scored against Coval's standard STT references.

GPT Realtime Whisper 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
OpenAI
Hosted by
OpenAI
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Multilingual

How GPT Realtime Whisper 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 GPT Realtime Whisper highlighted.
  1. #1STT RT v564 ms
  2. #2Parakeet TDT 0.6B v370 ms
  3. #3STT 183 ms
  4. #4Nova 399 ms
  5. #5Nova 2101 ms
  6. #6Ink 2108 ms
  7. #7Scribe v2 Realtime120 ms
  8. #8Universal 3.5 Pro146 ms
  9. #9Defaultvia Azure155 ms
  10. #10Whisper Large v3180 ms
  11. #11Velma 2 STT Streaming191 ms
  12. #12Grok STT199 ms
  13. #13Pulse205 ms
  14. #14Defaultvia Speechmatics223 ms
  15. #15Defaultvia Gradium249 ms
  16. #16resonant-1288 ms
  17. #17Enhanced341 ms
  18. #18Voxtral Mini Transcribe Realtime 2602356 ms
  19. #19GPT Realtime Whisper559 ms
Show all 24 models
  1. #20GPT-4o mini Transcribe623 ms
  2. #21Solaria 1680 ms
  3. #22GPT-4o Transcribe742 ms
  4. #23Chirp 2811 ms
  5. #24Chirp 3813 ms
median of all models · 202 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1STT RT v5Soniox64 ms1533 ms29,322
2Parakeet TDT 0.6B v3Together AI70 ms1210 ms29,161
3STT 1Inworld AI83 ms1468 ms29,280
4Nova 3Deepgram99 ms1434 ms29,290
5Nova 2Deepgram101 ms1436 ms29,145
6Ink 2Cartesia108 ms1812 ms29,288
7Scribe v2 RealtimeElevenLabs120 ms2157 ms29,313
8Universal 3.5 ProAssemblyAI146 ms1030 ms29,280
9DefaultAzure155 ms1792 ms6,682
10Whisper Large v3Together AI180 ms1241 ms29,171
11Velma 2 STT StreamingModulate191 ms1576 ms17,895
12Grok STTxAI199 ms—29,322
13PulseSmallest205 ms2072 ms29,313
14DefaultSpeechmatics223 ms1473 ms29,323
15DefaultGradium249 ms1979 ms28,235
16resonant-1Reson8288 ms—17,066
17EnhancedSpeechmatics341 ms1536 ms29,322
18Voxtral Mini Transcribe Realtime 2602Mistral356 ms1828 ms29,175
19GPT Realtime WhisperOpenAI559 ms1823 ms29,286
20GPT-4o mini TranscribeOpenAI623 ms—29,318
21Solaria 1Gladia680 ms1703 ms26,356
22GPT-4o TranscribeOpenAI742 ms—29,318
23Chirp 2Google811 ms6000 ms29,237
24Chirp 3Google813 ms5999 ms29,319
—FluxDeepgram—1089 ms29,354
—Flux MultilingualDeepgram—1175 ms29,344
—Nemotron 3.5 ASR StreamingTogether AI—1540 ms29,193
—Universal StreamingAssemblyAI—1510 ms29,297
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: 11th of 28 on Word Error Rate.

Latency vs accuracy

TTFS vs WEREach point is one measured model · GPT Realtime Whisper highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with GPT Realtime Whisper highlighted.
STT RT v5 — 64 ms, 5.8%Parakeet TDT 0.6B v3 — 70 ms, 11.0%STT 1 — 83 ms, 4.8%Nova 3 — 99 ms, 6.3%Nova 2 — 101 ms, 7.9%Ink 2 — 108 ms, 4.9%Scribe v2 Realtime — 120 ms, 5.5%Universal 3.5 Pro — 146 ms, 3.2%Default via Azure — 155 ms, 5.3%Whisper Large v3 — 180 ms, 8.2%Velma 2 STT Streaming — 191 ms, 6.4%Grok STT — 199 ms, 4.8%Pulse — 205 ms, 5.3%Default via Speechmatics — 223 ms, 5.5%Default via Gradium — 249 ms, 10.0%resonant-1 — 288 ms, 3.4%Enhanced — 341 ms, 4.3%Voxtral Mini Transcribe Realtime 2602 — 356 ms, 6.2%GPT-4o mini Transcribe — 623 ms, 5.1%Solaria 1 — 680 ms, 7.1%GPT-4o Transcribe — 742 ms, 4.7%Chirp 2 — 811 ms, 5.0%Chirp 3 — 813 ms, 4.1%GPT Realtime Whisper — 559 ms, 5.2%

Where the errors come from

WER compositionGPT Realtime Whisper's Word Error Rate split by error type · 30-day averageGPT Realtime Whisper's WER split into substitutions, deletions and insertions.
  • GPT Realtime Whisper5.2%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

GPT Realtime Whisper latency distributionMilliseconds · shared axis across metrics · last 30 daysGPT Realtime Whisper's latency percentiles per metric: p25–p75 band, p50 tick, whisker to p99.
  • Time to Final Segmentp50 560 ms · p99 952 ms
  • Time to First Tokenp50 1842 ms · p99 3632 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 Final Segment559 ms463 ms560 ms611 ms655 ms689 ms952 ms29,286
Word Error Rate5.2%0.0%0.0%6.5%15.4%23.1%52.4%29,327
Time to First Token1823 ms1525 ms1842 ms2039 ms2318 ms2623 ms3632 ms29,327

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 daysGPT Realtime Whisper's daily median Time to Final Segment over the last 30 days.
GPT Realtime Whisper · Jul 30: 584 msGPT Realtime Whisper · Jul 31: 591 msGPT Realtime Whisper · Aug 1: 547 msGPT Realtime Whisper · Aug 2: 554 msGPT Realtime Whisper · Aug 3: 515 msGPT Realtime Whisper · Aug 4: 568 msGPT Realtime Whisper · Aug 5: 556 msGPT Realtime Whisper · Aug 6: 573 msGPT Realtime Whisper · Aug 7: 575 msGPT Realtime Whisper · Aug 8: 502 msGPT Realtime Whisper · Aug 9: 541 msGPT Realtime Whisper · Aug 10: 584 msGPT Realtime Whisper · Aug 11: 534 msGPT Realtime Whisper · Aug 12: 584 msGPT Realtime Whisper · Aug 13: 523 msGPT Realtime Whisper · Aug 14: 539 msGPT Realtime Whisper · Aug 15: 542 msGPT Realtime Whisper · Aug 16: 520 msGPT Realtime Whisper · Aug 17: 562 msGPT Realtime Whisper · Aug 18: 554 msGPT Realtime Whisper · Aug 19: 510 msGPT Realtime Whisper · Aug 20: 578 msGPT Realtime Whisper · Aug 21: 555 msGPT Realtime Whisper · Aug 22: 525 msGPT Realtime Whisper · Aug 23: 499 msGPT Realtime Whisper · Aug 24: 571 msGPT Realtime Whisper · Aug 25: 597 msGPT Realtime Whisper · Aug 26: 528 msGPT Realtime Whisper · Aug 27: 533 msGPT Realtime Whisper · Aug 28: 580 ms
Word Error Rate — daily averageDaily average · UTC daysGPT Realtime Whisper's daily Word Error Rate over the last 30 days.
GPT Realtime Whisper · Jul 30: 4.9%GPT Realtime Whisper · Jul 31: 7.0%GPT Realtime Whisper · Aug 1: 5.8%GPT Realtime Whisper · Aug 2: 6.5%GPT Realtime Whisper · Aug 3: 4.8%GPT Realtime Whisper · Aug 4: 5.9%GPT Realtime Whisper · Aug 5: 7.7%GPT Realtime Whisper · Aug 6: 5.4%GPT Realtime Whisper · Aug 7: 7.5%GPT Realtime Whisper · Aug 8: 7.2%GPT Realtime Whisper · Aug 9: 5.3%GPT Realtime Whisper · Aug 10: 6.5%GPT Realtime Whisper · Aug 11: 7.0%GPT Realtime Whisper · Aug 12: 4.6%GPT Realtime Whisper · Aug 13: 6.1%GPT Realtime Whisper · Aug 14: 6.7%GPT Realtime Whisper · Aug 15: 5.4%GPT Realtime Whisper · Aug 16: 5.8%GPT Realtime Whisper · Aug 17: 6.0%GPT Realtime Whisper · Aug 18: 6.9%GPT Realtime Whisper · Aug 19: 6.8%GPT Realtime Whisper · Aug 20: 6.1%GPT Realtime Whisper · Aug 21: 4.8%GPT Realtime Whisper · Aug 22: 8.5%GPT Realtime Whisper · Aug 23: 7.4%GPT Realtime Whisper · Aug 24: 4.9%GPT Realtime Whisper · Aug 25: 6.9%GPT Realtime Whisper · Aug 26: 4.5%GPT Realtime Whisper · Aug 27: 5.8%GPT Realtime Whisper · Aug 28: 7.4%

Time to Final Segment by dataset

GPT Realtime Whisper by test conditionMilliseconds · lower is better · best condition firstGPT Realtime Whisper's Time to Final Segment on each benchmark dataset.
  1. WildASR noise gaps551 ms
  2. WildASR clean553 ms
  3. WildASR far-field559 ms
  4. WildASR phone codec559 ms
  5. PipeCat (production)560 ms
  6. WildASR clipping566 ms
  7. WildASR reverb567 ms
  8. WildASR accents567 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
LibriSpeech457 ms1
ProductionPipeCat560 ms14,932
AccentsWildASR567 ms1,438
CleanWildASR553 ms5,777
ClippingWildASR566 ms1,444
Far-fieldWildASR559 ms1,442
Noise gapsWildASR551 ms1,440
Phone codecWildASR559 ms1,440
ReverbWildASR567 ms1,372

Strongest condition: LibriSpeech at 457 ms · weakest: WildASR accents at 567 ms.

Limits of this comparison

This path is reported separately because session events and integration behavior differ from standalone audio-transcription requests.

  • The name describes Coval's integration path, and OpenAI can update Realtime defaults independently of standalone Whisper endpoints.

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