SpaceXAI says its new transcription model cuts short-phrase errors to 6.8%
single source· 1 articles · confidence: medium · first seen 2026-09-19 03:42 UTC
What this means for you
Nothing to act on unless accuracy was the thing keeping you on 1.0. The price has not moved, the model is live in the Grok Voice API, and the accuracy claim is the vendor's own — run it against your own audio before migrating.
SpaceXAI has shipped Grok Voice Transcribe 2.0, a speech-to-text model for batch and streaming audio, at the same prices as version 1.0: $0.10 an hour for batch, $0.20 an hour for streaming. The company says short-phrase word error rate (the share of words transcribed incorrectly) across 19 languages fell from 20.6% to 6.8%. No independent evaluation, evaluation date or test harness was given, so those figures are SpaceXAI's own. The model is available now through the Grok Voice API.
Key facts
- ·SpaceXAI released Grok Voice Transcribe 2.0, a speech-to-text model handling batch and streaming audio, available through the Grok Voice API. source
- ·Pricing is unchanged from version 1.0: $0.10 per hour for batch transcription and $0.20 per hour for streaming. source
- ·SpaceXAI reports short-phrase word error rate across 19 languages falling from 20.6% to 6.8%. source
- ·The company claims version 2.0 is twice as accurate as version 1.0 at the same price. source
What the sources say
- MarkTechPost — Reports the launch, the claimed error-rate drop across 19 languages and unchanged per-hour pricing.
Sources
The original reporting. Follow these — they did the work.