SpaceXAI says Grok Voice Transcribe 2 doubles accuracy
SpaceXAI is keeping Grok Voice Transcribe 2.0 at the old price while claiming a twofold accuracy gain.
The company says the new speech-to-text model ranks first among 32 streaming models on Artificial Analysis’ public accuracy leaderboard. Its reported gains include telephony audio, English support calls, spoken credentials, short commands in 19 languages, and multilingual short phrases. The API adds or keeps support for real-time streams, batch jobs, diarization, timestamps, confidence scores, multi-channel transcription, and biasing for specialized terms. Existing integrations get the update without code changes, though teams can temporarily pin version 1.0 before it is deprecated in the coming weeks. TestingCatalog's note
The company says the new speech-to-text model ranks first among 32 streaming models on Artificial Analysis’ public accuracy leaderboard. Its reported gains include telephony audio, English support calls, spoken credentials, short commands in 19 languages, and multilingual short phrases. The API adds or keeps support for real-time streams, batch jobs, diarization, timestamps, confidence scores, multi-channel transcription, and biasing for specialized terms. Existing integrations get the update without code changes, though teams can temporarily pin version 1.0 before it is deprecated in the coming weeks. TestingCatalog's note
score 5