SPEECHIFYAITEXT-TO-SPEECH7,523 SAMPLES / 30 DAYSLAST RUN NOV 21, 2026, 03:30 UTC

Simba 3.5 text-to-speech benchmarks

Simba 3.5, hosted by SpeechifyAI, measures mean 114 ms time to first audio (10th of 32) and 2.4% word error rate (30th of 32) among TTS systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Coval has active text-to-speech measurements for Simba 3.5, created by SpeechifyAI and served through SpeechifyAI's own API.

Time to First Audio#10 / 32
114ms
TTFA Network Roundtrip#10 / 32
103ms
TTFA Leading Silence#9 / 32
11ms
Word Error Rate#30 / 32
2.4%

Overview

Simba 3.5 is tested every day on a fixed set of text prompts, measuring how quickly audible speech starts and how intelligible the result is.

Technical specifications

Hosted by
SpeechifyAI
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Emotion control, Multilingual, Voice cloning

How Simba 3.5 ranks

Full TTS dashboard
Time to First Audio — every measured TTS modelMilliseconds · lower is better · 30-day averageEvery TTS model ranked on Time to First Audio, with Simba 3.5 highlighted.
  1. #1vui40 ms
  2. #2Default62 ms
  3. #3PhantomZ 4.064 ms
  4. #4TTS Flash 268 ms
  5. #6Simba 3.2101 ms
  6. #7Qwen3 TTS 1.7b105 ms
  7. #8Palabra TTS107 ms
  8. #9Simba 3.0113 ms
  9. #10Simba 3.5114 ms
  10. #11Falcon 2139 ms
  11. #12TTS 2163 ms
Show all 32 models
  1. #13Flash v2.5215 ms
  2. #14Flux TTS218 ms
  3. #17Coda254 ms
  4. #18Airy TTS273 ms
  5. #19Sonic 3.6287 ms
  6. #20TTS Rt v2288 ms
  7. #21TTS RT v1288 ms
  8. #23Grok TTS311 ms
  9. #24Aura 2313 ms
  10. #26Daytona TTS347 ms
  11. #27S2.1 Pro383 ms
  12. #28Mist v3397 ms
  13. #29Aura 2 En515 ms
  14. #30Chirp 3 HD569 ms
  15. #32S2.1 Pro Free842 ms
median of all models · 237 ms
TTS models over the last 30 days, ranked on Time to First Audio.
#ModelHostTTFAWERSamples
1vuiFluxions40 ms2,870
2Gradium TTS Beta 202609Gradium62 ms2,876
3Phantom Z 3.4 conversationalDeepdub64 ms2,872
4TTS Flash 2Inworld AI68 ms2,876
5Qwen3 TTS FastNari Labs69 ms2,876
6Simba 3.2SpeechifyAI101 ms2,875
7Qwen3 TTS 1.7bBaseten105 ms2,875
8Palabra TTS v1Palabra107 ms2,469
9Simba 3.0SpeechifyAI113 ms2,876
10Simba 3.5SpeechifyAI114 ms1,759
11Falcon 2Murf139 ms2,875
12TTS 2Inworld AI163 ms2,874
13Flash v2.5ElevenLabs215 ms2,872
14Flux TTSDeepgram218 ms2,378
15Eleven v4 TurboElevenLabs218 ms2,875
16Assemblyai TTS 1AssemblyAI220 ms2,526
17CodaRime254 ms1,882
18Airy TTSOut of Set273 ms2,789
19Sonic 3.6Cartesia287 ms1,933
20TTS Rt v2Soniox288 ms2,379
21TTS RT v1Soniox288 ms2,395
22Drama 3 Preview Us EastFish Audio293 ms2,875
23Grok TTSSpaceXAI311 ms2,876
24Aura 2Deepgram313 ms2,529
25Lightning v3.1 ProSmallest327 ms2,868
26Daytona TTSGuava347 ms2,880
27S2.1 ProFish Audio383 ms2,871
28Mist v3Rime397 ms1,882
29Aura 2 EnCloudflare515 ms2,866
30Chirp 3 HDGoogle569 ms2,875
31Qwen3 TTS Flash RealtimeAlibaba722 ms2,875
32S2.1 Pro FreeFish Audio842 ms2,857
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

Where the errors come from

WER compositionSimba 3.5's Word Error Rate split by error type · 30-day averageSimba 3.5's WER split into substitutions, deletions and insertions.
  • Simba 3.52.4%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Simba 3.5 latency distributionMilliseconds · shared axis across metrics · last 30 daysSimba 3.5's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to First Audiop50 106 ms · Max 4072 ms
  • TTFA Network Roundtripp50 96 ms · Max 4061 ms
  • TTFA Leading Silencep50 11 ms · Max 17 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 First Audio114 ms91 ms106 ms125 ms148 ms163 ms4072 ms1,759
TTFA Network Roundtrip103 ms80 ms96 ms114 ms137 ms152 ms4061 ms1,759
TTFA Leading Silence11 ms8 ms11 ms13 ms14 ms15 ms17 ms1,759
Word Error Rate2.4%0.0%0.0%4.3%10.7%18.8%43.8%1,759

Last 30 days

Daily medians from the same measurement runs · gaps are days without qualifying runs.

Time to First Audio — daily p50Line p50 · band p25–p75 · UTC daysSimba 3.5's daily median Time to First Audio over the last 30 days.
Simba 3.5 · Nov 2: 129 msSimba 3.5 · Nov 3: 119 msSimba 3.5 · Nov 4: 110 msSimba 3.5 · Nov 5: 125 msSimba 3.5 · Nov 6: 137 msSimba 3.5 · Nov 7: 124 msSimba 3.5 · Nov 8: 122 msSimba 3.5 · Nov 9: 120 msSimba 3.5 · Nov 10: 127 msSimba 3.5 · Nov 11: 123 msSimba 3.5 · Nov 12: 143 msSimba 3.5 · Nov 13: 90 msSimba 3.5 · Nov 14: 95 msSimba 3.5 · Nov 15: 87 msSimba 3.5 · Nov 16: 92 msSimba 3.5 · Nov 17: 95 msSimba 3.5 · Nov 18: 97 msSimba 3.5 · Nov 19: 96 msSimba 3.5 · Nov 20: 92 msSimba 3.5 · Nov 21: 89 ms
Word Error Rate — dailyDaily corpus-level · UTC daysSimba 3.5's daily Word Error Rate over the last 30 days.
Simba 3.5 · Nov 2: 4.7%Simba 3.5 · Nov 3: 4.2%Simba 3.5 · Nov 4: 4.9%Simba 3.5 · Nov 5: 3.1%Simba 3.5 · Nov 6: 3.6%Simba 3.5 · Nov 7: 3.4%Simba 3.5 · Nov 8: 3.5%Simba 3.5 · Nov 9: 3.4%Simba 3.5 · Nov 10: 3.5%Simba 3.5 · Nov 11: 4.9%Simba 3.5 · Nov 12: 2.5%Simba 3.5 · Nov 13: 3.5%Simba 3.5 · Nov 14: 2.1%Simba 3.5 · Nov 15: 4.0%Simba 3.5 · Nov 16: 4.0%Simba 3.5 · Nov 17: 3.1%Simba 3.5 · Nov 18: 4.6%Simba 3.5 · Nov 19: 5.3%Simba 3.5 · Nov 20: 3.4%Simba 3.5 · Nov 21: 4.1%

How fast is Simba 3.5?

On SpeechifyAI, Simba 3.5 measures mean 114 ms time to first audio (10th of 32). Last measured 2026-11-21.

How accurate is Simba 3.5?

On SpeechifyAI, Simba 3.5 measures 2.4% word error rate (30th of 32). Last measured 2026-11-21.

Who hosts Simba 3.5?

Simba 3.5 is created by SpeechifyAI and served by SpeechifyAI. 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