Announcement
SOV-Transcribe-1 tops the open-weight speech-to-text ranking, says @ArtificialAnlys
With a 2.3% AA-WER Index score, SOV-Transcribe-1 ranks seventh of 44 models overall, according to @ArtificialAnlys.
TLDR
@ArtificialAnlys ranks SOV-Transcribe-1 first among open-weight speech-to-text models and seventh among 44 overall, including proprietary models. It reports a 2.3% AA-WER Index score and 4.6 billion active parameters. @SFResearch pitches the model for accents, noisy rooms and specialized terminology, turning speech into something AI agents can act on.
Combined views
—
2 Sources, first seen ago
Comments