xAI Launches Grok Voice Transcribe 2.0, Doubling Accuracy

SAN FRANCISCO — xAI has released Grok Voice Transcribe 2.0, a new speech-to-text model the company says is twice as accurate as its predecessor — and it comes at the same price. The launch cements xAI's push to make Grok not just a chatbot, but the voice layer running underneath a growing range of real-world products.

Best-in-class accuracy

xAI says Grok Voice Transcribe 2.0 ranks first for accuracy among 32 streaming models on the public Artificial Analysis leaderboard. The company detailed the release on its site, noting the model was trained on live, noisy, multilingual audio recorded across a wide range of environments — exactly the messy conditions real transcription has to handle.

The gains are dramatic where they matter most. On a set of short multilingual voice commands spanning 19 languages — the kind of clipped phrases that trip up most systems — word error rate fell from 20.6% to just 6.8%. Multilingual transcription overall saw the largest improvement, with the model able to detect languages automatically and follow mid-sentence language switches in a single pass.

Powering products people already use

Grok Voice already powers tens of thousands of customer-support calls a day, transcribes millions of hours of video narration, and runs voice agents in physical products — including the Grok assistant now rolling out across Tesla vehicles and other platforms. Version 2.0 upgrades all of those experiences instantly, with xAI noting that existing Speech-to-Text API integrations get the accuracy boost with no code changes at all.

The feature set is deep: word-level timestamps with confidence scores, speaker diarization at no extra cost, multichannel transcription of up to eight channels, custom key-term biasing, and smart turn detection for voice agents. Atlassian has already adopted the model to transcribe every video on Loom, its screen-recording product.

Part of a relentless release cadence

The launch fits a pattern of rapid-fire shipping at xAI, now operating as part of the SpaceX-xAI combination. The company has been iterating aggressively across its lineup, from voice models to its flagship chatbot, as it races toward the next major Grok releases Musk has mapped out.

Crucially, xAI held pricing flat at $0.10 per hour of audio for batch transcription and $0.20 per hour for streaming, with diarization, timestamps, and key terms all included. Doubling accuracy while keeping the price unchanged is the kind of move that pressures rivals and rewards developers — and it signals that xAI intends to compete hard for the fast-growing voice-AI market. Version 2.0 will soon become the default in the Speech-to-Text API, putting best-in-class transcription in front of every Grok developer and, by extension, the millions of people who talk to Grok every day.