Grok Voice Transcribe 2.0 launches

SpaceXAI released Grok Voice Transcribe 2.0, a speech-to-text model that's twice as accurate as version 1.0 while costing the same. It ranks first for accuracy among 32 streaming models on the Artificial Analysis leaderboard.

“twice as accurate as Grok Voice Transcribe 1.0, at the same price”

Built for real-world audio

Grok Voice Transcribe 2.0 handles the messiest audio: phone lines, overlapping voices, accents, and spoken credentials. It's trained on live, noisy, multilingual recordings from diverse environments.

Multilingual breakthrough

Grok Voice Transcribe 2.0 transcribes dozens of languages, auto-detects the language, and handles mid-recording switches in one pass. On short phrases, word error rate dropped from 20.6% to 6.8%.

Existing integrations get better

Current Speech-to-Text API users get the accuracy boost with no code changes. The model supports batch and streaming, word-level timestamps, speaker diarization, multichannel transcription, and key term biasing.

Atlassian Loom partnership

Atlassian Loom now uses Grok Voice Transcribe 2.0 for video transcription. The accurate transcripts enable new workflows: record an action plan in Loom, pipe it to Cursor, and it generates code updates automatically.

“record what you mean, and the work gets done”

Same price, better accuracy

Batch transcription costs $0.10 per hour of audio, streaming $0.20 per hour. Diarization, timestamps, and key term biasing are included at no extra cost.

Version 1.0 being deprecated

Grok Voice Transcribe 2.0 will become the default in the Speech-to-Text API. Version 1.0 will be deprecated in the coming weeks, but users can pin it to stay on the older model during transition.