Speech-to-Text · Multilingual

Norwegian

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.

Sonioxstt-rt-v5Try →8.0%
ElevenLabsScribe v2 RealtimeTry →8.3%
AssemblyAIAssemblyAIUniversal-3.5 ProTry →12.3%
ModulateModulateVelma 2 multilingualTry →13.0%
AlibabaQwen3-ASR Realtime14.5%
DeepgramNova-3Try →14.9%
DeepgramNova-2Try →18.6%
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.

DeepgramNova-3Try →326ms · 5.35s
DeepgramNova-2Try →336ms · 5.27s
Sonioxstt-rt-v5Try →521ms · 710ms
AssemblyAIAssemblyAIUniversal-3.5 ProTry →566ms · 1.43s
ModulateModulateVelma 2 multilingualTry →577ms · 1.24s
OpenAIGPT TranscribeTry →854ms · 1.07s
OpenAIGPT-4o mini TranscribeTry →867ms · 1.02s
OpenAIGPT-4o TranscribeTry →905ms · 1.24s
AlibabaQwen3-ASR Realtime915ms · 1.04s
ElevenLabsScribe v2 RealtimeTry →1.92s · 2.13s
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 →2.16s · 2.77s
DeepgramNova-3Try →2.22s · 3.23s
DeepgramNova-2Try →2.23s · 3.22s
Sonioxstt-rt-v5Try →2.81s · 3.43s
ModulateModulateVelma 2 multilingualTry →2.92s · 3.34s
ElevenLabsScribe v2 RealtimeTry →3.15s · 3.21s
OpenAIGPT-4o mini TranscribeTry →7.54s · 11.1s
OpenAIGPT-4o TranscribeTry →7.55s · 11.2s
OpenAIGPT TranscribeTry →7.98s · 11.3s
AlibabaQwen3-ASR Realtime8.17s · 8.80s
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