Start with 45 free minutes of callsTry for free

Speech-to-Text · Multilingual

Japanese

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y
20ms50ms100ms200ms500ms1.00s2.00s2.0%5.0%10.0%20.0%Finalize latency lower is better · log scaleCER lower is better · log scaleSoniox stt-rt-v5Soniox stt-rt-v6OpenAI GPT-4o TranscribeOpenAI GPT TranscribeOpenAI GPT-4o mini TranscribeDeepgram Nova-3Modulate Velma 2 multilingualAssemblyAI Universal-3.5 ProElevenLabs Scribe v2 RealtimexAI Grok STTspeko.ai/benchmarks

CER

Character error rate — word scoring mismeasures this language.

Sonioxstt-rt-v5Try →2.2%
Sonioxstt-rt-v6Try →2.3%
ModulateModulateVelma 2 multilingualTry →3.0%
ElevenLabsScribe v2 RealtimeTry →4.2%
AssemblyAIAssemblyAIUniversal-3.5 ProTry →4.5%
OpenAIGPT-4o mini TranscribeTry →4.7%
OpenAIGPT-4o TranscribeTry →4.7%
OpenAIGPT TranscribeTry →5.0%
OpenAIGPT Live TranscribeTry →5.5%
DeepgramNova-3Try →7.5%
GladiaSolaria-1Try →17.5%
xAIGrok STTTry →29.3%
CartesiaInk-WhisperTry →37.1%
0.0%10.0%20.0%30.0%40.0%50.0%

Finalize latency

lower is better

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

DeepgramNova-3Try →25ms · 309ms
Sonioxstt-rt-v5Try →111ms · 313ms
Sonioxstt-rt-v6Try →111ms · 298ms
AssemblyAIAssemblyAIUniversal-3.5 ProTry →463ms · 709ms
OpenAIGPT-4o mini TranscribeTry →718ms · 835ms
OpenAIGPT TranscribeTry →873ms · 1.11s
OpenAIGPT-4o TranscribeTry →901ms · 1.24s
xAIGrok STTTry →1.29s · 2.35s
ElevenLabsScribe v2 RealtimeTry →2.02s · 2.22s
ModulateModulateVelma 2 multilingualTry →2.93s · 4.89s
0ms300ms600ms900ms1.20s

p50 p90

Time to first token

lower is better

How long before the first word of the transcript arrives.

AssemblyAIAssemblyAIUniversal-3.5 ProTry →2.01s · 2.71s
GladiaSolaria-1Try →2.34s · 4.66s
Sonioxstt-rt-v5Try →2.52s · 3.58s
DeepgramNova-3Try →2.61s · 3.81s
xAIGrok STTTry →2.63s · 5.71s
Sonioxstt-rt-v6Try →2.84s · 3.92s
ElevenLabsScribe v2 RealtimeTry →3.10s · 4.20s
ModulateModulateVelma 2 multilingualTry →4.36s · 5.51s
CartesiaInk-WhisperTry →5.00s · 10.6s
OpenAIGPT-4o mini TranscribeTry →5.25s · 8.30s
OpenAIGPT-4o TranscribeTry →5.84s · 10.6s
OpenAIGPT TranscribeTry →5.89s · 10.9s
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