Voice AI BenchmarksBy

DEEPGRAMSPEECH-TO-TEXT113,951 SAMPLES / 30 DAYSLAST RUN SEP 15, 2026, 08:30 UTC

official resource

Nova 2 speech-to-text benchmarks

Nova 2, hosted by Deepgram, measures mean 92 ms time to final segment (7th of 28) and 7.9% word error rate (25th of 30) among STT systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Nova 2 is Deepgram's previous-generation general speech recognition model.

Time to Final Segment#7 / 28
92ms
Word Error Rate#25 / 30
7.9%
Time to First Token#10 / 26
1420ms

Overview

Deepgram serves Nova 2 through its streaming API with multilingual, diarization, code-switching and keyterm options.

Nova 2 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
Deepgram
Hosted by
Deepgram
Source
Official API
Licensing
Proprietary
Deployment
On-prem
Region
US
Features
Code switching, Diarization, Keyterm biasing, Multilingual, VAD

How Nova 2 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 Nova 2 highlighted.
  1. #1Qwen3 ASR 1.7b39 ms
  2. #2Qwen3 ASR Fast46 ms
  3. #3STT RT v557 ms
  4. #4STT 165 ms
  5. #5Parakeet TDT 0.6B v381 ms
  6. #6Nova 389 ms
  7. #7Nova 292 ms
  8. #8Flux Multilingual98 ms
  9. #9Flux99 ms
  10. #10Ink 2122 ms
  11. #11Whisper Large v3via Baseten125 ms
  12. #12Scribe v2 Realtime133 ms
Show all 28 models
  1. #13Universal 3.5 Pro173 ms
  2. #14Grok STT207 ms
  3. #15Defaultvia Speechmatics209 ms
  4. #16Pulse212 ms
  5. #17Defaultvia Gradium246 ms
  6. #18resonant-1264 ms
  7. #19Whisper Large v3via Together AI295 ms
  8. #20Enhanced299 ms
  9. #21Gemini 3.5 Transcribe Live305 ms
  10. #22Voxtral Mini Transcribe Realtime 2602401 ms
  11. #23GPT Realtime Whisper551 ms
  12. #24GPT-4o mini Transcribe690 ms
  13. #25GPT-4o Transcribe754 ms
  14. #26Chirp 3776 ms
  15. #27Solaria 1805 ms
  16. #28Chirp 2873 ms
median of all models · 208 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1Qwen3 ASR 1.7bBaseten39 ms931 ms1,256
2Qwen3 ASR FastNari46 ms1748 ms4,451
3STT RT v5Soniox57 ms1529 ms22,468
4STT 1Inworld AI65 ms1400 ms22,836
5Parakeet TDT 0.6B v3Together AI81 ms1215 ms22,488
6Nova 3Deepgram89 ms1419 ms22,855
7Nova 2Deepgram92 ms1420 ms22,733
8Flux MultilingualDeepgram98 ms1164 ms12,251
9FluxDeepgram99 ms1076 ms12,243
10Ink 2Cartesia122 ms1827 ms22,862
11Whisper Large v3Baseten125 ms912 ms1,252
12Scribe v2 RealtimeElevenLabs133 ms2175 ms22,866
13Universal 3.5 ProAssemblyAI173 ms1041 ms20,253
14Grok STTxAI207 ms—22,853
15DefaultSpeechmatics209 ms1428 ms22,876
16PulseSmallest212 ms2011 ms22,856
17DefaultGradium246 ms1980 ms22,839
18resonant-1Reson8264 ms—22,860
19Whisper Large v3Together AI295 ms1338 ms22,744
20EnhancedSpeechmatics299 ms1492 ms22,876
21Gemini 3.5 Transcribe LiveGemini305 ms1679 ms16,403
22Voxtral Mini Transcribe Realtime 2602Mistral401 ms1851 ms22,583
23GPT Realtime WhisperOpenAI551 ms1814 ms22,812
24GPT-4o mini TranscribeOpenAI690 ms—22,830
25GPT-4o TranscribeOpenAI754 ms—22,831
26Chirp 3Google776 ms5974 ms22,875
27Solaria 1Gladia805 ms1804 ms22,439
28Chirp 2Google873 ms6071 ms22,789
—Nemotron 3.5 ASR StreamingTogether AI—1548 ms22,747
—Universal StreamingAssemblyAI—1514 ms20,235
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 · Nova 2 highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with Nova 2 highlighted.
Qwen3 ASR 1.7b — 39 ms, 4.1%Qwen3 ASR Fast — 46 ms, 3.2%STT RT v5 — 57 ms, 5.7%STT 1 — 65 ms, 4.4%Parakeet TDT 0.6B v3 — 81 ms, 11.2%Nova 3 — 89 ms, 6.2%Flux Multilingual — 98 ms, 8.8%Flux — 99 ms, 6.7%Ink 2 — 122 ms, 4.8%Whisper Large v3 via Baseten — 125 ms, 5.6%Scribe v2 Realtime — 133 ms, 5.5%Universal 3.5 Pro — 173 ms, 3.1%Grok STT — 207 ms, 4.6%Default via Speechmatics — 209 ms, 5.4%Pulse — 212 ms, 5.3%Default via Gradium — 246 ms, 10.0%resonant-1 — 264 ms, 3.4%Whisper Large v3 via Together AI — 295 ms, 8.5%Enhanced — 299 ms, 4.3%Gemini 3.5 Transcribe Live — 305 ms, 3.9%Voxtral Mini Transcribe Realtime 2602 — 401 ms, 6.2%GPT Realtime Whisper — 551 ms, 5.1%GPT-4o mini Transcribe — 690 ms, 4.9%GPT-4o Transcribe — 754 ms, 4.6%Chirp 3 — 776 ms, 4.1%Solaria 1 — 805 ms, 7.1%Chirp 2 — 873 ms, 4.9%Nova 2 — 92 ms, 7.9%

Where the errors come from

WER compositionNova 2's Word Error Rate split by error type · 30-day averageNova 2's WER split into substitutions, deletions and insertions.
  • Nova 27.9%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Nova 2 latency distributionMilliseconds · shared axis across metrics · last 30 daysNova 2's latency percentiles per metric: p25–p75 band, p50 tick, whisker to p99.
  • Time to Final Segmentp50 92 ms · p99 193 ms
  • Time to First Tokenp50 993 ms · p99 2994 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 Segment92 ms59 ms92 ms104 ms126 ms143 ms193 ms22,733
Word Error Rate7.9%0.0%3.8%10.5%20.0%28.6%68.4%22,864
Time to First Token1420 ms981 ms993 ms1981 ms1995 ms2028 ms2994 ms22,864

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 daysNova 2's daily median Time to Final Segment over the last 30 days.
Nova 2 · Sep 1: 89 msNova 2 · Sep 2: 89 msNova 2 · Sep 3: 93 msNova 2 · Sep 4: 94 msNova 2 · Sep 5: 87 msNova 2 · Sep 6: 87 msNova 2 · Sep 7: 80 msNova 2 · Sep 8: 92 msNova 2 · Sep 9: 95 msNova 2 · Sep 10: 90 msNova 2 · Sep 11: 95 msNova 2 · Sep 12: 79 msNova 2 · Sep 13: 87 msNova 2 · Sep 14: 91 msNova 2 · Sep 15: 89 ms
Word Error Rate — daily averageDaily average · UTC daysNova 2's daily Word Error Rate over the last 30 days.
Nova 2 · Sep 1: 9.7%Nova 2 · Sep 2: 8.5%Nova 2 · Sep 3: 8.7%Nova 2 · Sep 4: 7.7%Nova 2 · Sep 5: 8.2%Nova 2 · Sep 6: 8.5%Nova 2 · Sep 7: 9.5%Nova 2 · Sep 8: 7.7%Nova 2 · Sep 9: 8.4%Nova 2 · Sep 10: 8.0%Nova 2 · Sep 11: 8.8%Nova 2 · Sep 12: 8.1%Nova 2 · Sep 13: 9.3%Nova 2 · Sep 14: 9.2%Nova 2 · Sep 15: 8.7%

Time to Final Segment by dataset

Nova 2 by test conditionMilliseconds · lower is better · best condition firstNova 2's Time to Final Segment on each benchmark dataset.
  1. WildASR reverb86 ms
  2. WildASR noise gaps87 ms
  3. WildASR clipping87 ms
  4. WildASR accents90 ms
  5. WildASR clean90 ms
  6. PipeCat (production)91 ms
  7. WildASR phone codec93 ms
  8. WildASR far-field126 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
LibriSpeech106 ms1
ProductionPipeCat91 ms11,505
AccentsWildASR90 ms1,116
CleanWildASR90 ms4,578
ClippingWildASR87 ms1,120
Far-fieldWildASR126 ms1,136
Noise gapsWildASR87 ms1,143
Phone codecWildASR93 ms1,135
ReverbWildASR86 ms999

Strongest condition: WildASR reverb at 86 ms · weakest: WildASR far-field at 126 ms.

How fast is Nova 2?

On Deepgram, Nova 2 measures mean 92 ms time to final segment (7th of 28) and mean 1420 ms time to first token (10th of 26). Last measured 2026-09-15.

How accurate is Nova 2?

On Deepgram, Nova 2 measures 7.9% word error rate (25th of 30). Last measured 2026-09-15.

Who hosts Nova 2?

Nova 2 is created by Deepgram and served by Deepgram. Coval measures each hosted endpoint separately.

Limits of this comparison

Keeping it active provides a consistent baseline for judging Nova 3 without comparing results from different audio.

  • One general benchmark configuration cannot represent every Nova 2 language, domain variant or formatting option.

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