Start building →OverviewMultilingualCode-switchingEmotion

Speech-to-Text · Multilingual

Vietnamese

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y
50ms100ms200ms500ms1.00s2.00s5.0%10.0%Finalize latency lower is better · log scaleWER lower is betterOpenAI GPT-4o TranscribeOpenAI GPT TranscribeGoogle Gemini 3.5 Transcribe LiveSoniox stt-rt-v5Alibaba Qwen3-ASR RealtimeAssemblyAI Universal-3.5 ProOpenAI GPT-4o mini TranscribeOpenAI GPT Live TranscribexAI Grok STTMeta Muse Voice TranscribeNari Qwen3-ASRNari Qwen3-ASR FastElevenLabs Scribe v2 RealtimeDeepgram Nova-3Gladia Solaria-1Cartesia Ink-Whisper ↓ 16.7%speko.ai/benchmarks

WER

Share of words the transcript got wrong.

GoogleGemini 3.5 Transcribe LiveTry →1.9%
Sonioxstt-rt-v5Try →2.3%
AlibabaQwen3-ASR RealtimeTry →2.5%
OpenAIGPT Live TranscribeTry →2.8%
NariNariQwen3-ASRTry →3.5%
NariNariQwen3-ASR FastTry →3.6%
OpenAIGPT Realtime Whisper3.7%
MetaMetaMuse Voice TranscribeTry →3.7%
xAIGrok STTTry →3.9%
OpenAIGPT TranscribeTry →4.3%
ElevenLabsScribe v2 RealtimeTry →4.4%
DeepgramNova-3Try →6.1%
OpenAIGPT-4o TranscribeTry →7.9%
AssemblyAIAssemblyAIUniversal-3.5 ProTry →8.3%
GladiaSolaria-1Try →9.5%
OpenAIGPT-4o mini TranscribeTry →10.6%
CartesiaInk-WhisperTry →16.7%
0.0%5.0%10.0%15.0%20.0%25.0%

Finalize latency

Region

Time from the caller stopping to the finalized transcript, not the first partial — the agent cannot reply until it lands.

NariNariQwen3-ASRTry →34ms · 56ms
NariNariQwen3-ASR FastTry →39ms · 59ms
DeepgramNova-3Try →56ms · 178ms
MetaMetaMuse Voice TranscribeTry →241ms · 397ms
AssemblyAIAssemblyAIUniversal-3.5 ProTry →242ms · 534ms
Sonioxstt-rt-v5Try →319ms · 1.45s
CartesiaInk-WhisperTry →323ms · 450ms
OpenAIGPT Live TranscribeTry →490ms · 615ms
OpenAIGPT TranscribeTry →711ms · 915ms
OpenAIGPT-4o TranscribeTry →713ms · 885ms
OpenAIGPT-4o mini TranscribeTry →719ms · 945ms
GladiaSolaria-1Try →724ms · 966ms
AlibabaQwen3-ASR RealtimeTry →869ms · 989ms
xAIGrok STTTry →980ms · 1.14s
GoogleGemini 3.5 Transcribe LiveTry →1.26s · 1.34s
ElevenLabsScribe v2 RealtimeTry →1.68s · 1.79s
0ms300ms600ms900ms1.20s

p50 p90

Time to first token

Region

How long before the first word of the transcript arrives.

AssemblyAIAssemblyAIUniversal-3.5 ProTry →1.10s · 1.81s
Sonioxstt-rt-v5Try →1.64s · 2.32s
GladiaSolaria-1Try →1.69s · 2.49s
NariNariQwen3-ASR FastTry →1.74s · 1.75s
NariNariQwen3-ASRTry →1.76s · 1.77s
MetaMetaMuse Voice TranscribeTry →1.80s · 2.52s
OpenAIGPT Live TranscribeTry →1.83s · 2.69s
DeepgramNova-3Try →1.87s · 2.03s
GoogleGemini 3.5 Transcribe LiveTry →1.90s · 2.57s
xAIGrok STTTry →2.06s · 2.15s
ElevenLabsScribe v2 RealtimeTry →2.22s · 3.02s
OpenAIGPT-4o TranscribeTry →3.87s · 6.25s
OpenAIGPT-4o mini TranscribeTry →3.89s · 6.34s
CartesiaInk-WhisperTry →4.01s · 6.05s
OpenAIGPT TranscribeTry →4.02s · 6.26s
AlibabaQwen3-ASR RealtimeTry →6.76s · 7.79s
0ms2.00s4.00s6.00s8.00s

p50 p90

Cost

lower is better

$ per 1,000 minutes of audio · vendor list price, the same in every language

$1.00
$2.00
$2.00
$2.10
$3.00
$3.70
$4.50
$4.80
$5.40
$6.00
$6.50
$7.50
$9.00
$12.50
$17.00
$17.00
NariNariQwen3-ASRNariNariQwen3-ASR FastSonioxstt-rt-v5AlibabaQwen3-ASRMetaMetaMuse Voice TranscribeElevenLabsScribe v2OpenAIGPT TranscribeDeepgramNova-3AlibabaQwen3-ASR RealtimeOpenAIGPT-4o TranscribeElevenLabsScribe v2 RealtimeAssemblyAIAssemblyAIUniversal-3.5 ProGoogleGemini 3.5 Transcribe LiveGladiaSolaria-1OpenAIGPT Live TranscribeOpenAIGPT Realtime Whisper