Speech-to-Text · Multilingual

German

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y

WER

Share of words the transcript got wrong.

AssemblyAIAssemblyAIUniversal-3.6 ProTry →3.11%
AssemblyAIAssemblyAIUniversal-3.5 ProTry →3.61%
ElevenLabsScribe v2 RealtimeTry →4.07%
ModulateModulateVelma 2 multilingualTry →4.49%
OpenAIGPT-4o TranscribeTry →4.5%
NariNariQwen3-ASRTry →4.58%
MetaMetaMuse Voice Transcribe 1.0Try →4.61%
Sonioxstt-rt-v5Try →4.99%
GoogleChirp 35.27%
GradiumASRTry →5.48%
AlibabaQwen3-ASR Realtime5.5%
OpenAIGPT-4o-mini TranscribeTry →6.31%
Smallest AISmallest AIPulseTry →9.73%
DeepgramNova-3Try →10.22%
DeepgramNova-2Try →10.97%
CartesiaInk-WhisperTry →11.03%
0.0%5.0%10.0%15.0%20.0%25.0%

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.

CartesiaInk-WhisperTry →313ms · 441ms
DeepgramNova-3Try →321ms · 396ms
DeepgramNova-2Try →328ms · 420ms
Sonioxstt-rt-v5Try →637ms · 2.68s
OpenAIGPT-4o-mini TranscribeTry →875ms · 1.14s
Smallest AISmallest AIPulseTry →891ms · 1.04s
AlibabaQwen3-ASR Realtime937ms · 1.05s
GradiumASRTry →989ms · 1.33s
OpenAIGPT-4o TranscribeTry →1.02s · 1.32s
GoogleChirp 31.19s · 1.63s
AssemblyAIAssemblyAIUniversal-3.5 ProTry →1.49s · 1.91s
ElevenLabsScribe v2 RealtimeTry →1.95s · 2.22s
0ms300ms600ms900ms1.20s

p50 p90

Time to first token

lower is better

How long before the first word of the transcript arrives.

DeepgramNova-3Try →2.12s · 5.21s
AssemblyAIAssemblyAIUniversal-3.5 ProTry →2.17s · 4.69s
DeepgramNova-2Try →2.23s · 5.95s
Sonioxstt-rt-v5Try →2.52s · 5.14s
GradiumASRTry →2.94s · 5.63s
ElevenLabsScribe v2 RealtimeTry →3.11s · 5.23s
Smallest AISmallest AIPulseTry →3.58s · 5.80s
CartesiaInk-WhisperTry →5.13s · 10.6s
OpenAIGPT-4o TranscribeTry →6.50s · 12.3s
OpenAIGPT-4o-mini TranscribeTry →6.51s · 12.3s
GoogleChirp 36.54s · 9.66s
AlibabaQwen3-ASR Realtime7.39s · 10.5s
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