xAI: Grok Voice Transcribe 2.0 takes the audio, priced by the hour
On 18 September 2026, xAI releases Grok Voice Transcribe 2.0. The page calls it twice as accurate as version 1.0, at the same price: 0.10 dollar per hour in batch, 0.20 streaming. The audio leaves for xAI. The accuracy is their measurement.
Getting started
Original AI-generated illustration · SecuFocus
At a glance
Key points
This is not a local recorder. It is a transcription model built on the same base as Grok Voice, already used, the page says, for support calls, hours of video, and the Grok assistant in Tesla vehicles.
What changes
The page is dated 18 September 2026. The model aims at real audio: a bad line, several voices, accents, numbers and addresses read aloud. Diarization, timestamps and key terms are announced as included in the price, not as an option.
“Twice as accurate” compares with Grok Voice Transcribe 1.0, on xAI’s evaluations. It is not a comparison we reran, or a ranking against other services.
What leaves
Sending an hour of audio to the API means sending the voices, the names spoken, the phone numbers and the background. The hourly price says nothing about how long it is kept. The page we read does not set that duration.
The same base, xAI says, already handles support calls and the assistant in Teslas. Connecting Transcribe 2.0 to a meeting recording is the same gesture: the meeting leaves the room.
What we did not verify
We did not send a file, or measure a transcription error. The doubling of accuracy is xAI’s sentence.
The decision
If the recording holds voices that did not agree to leave the room, do not send it. A low price does not make the send harmless. A local tool, if you need a trace that stays on disk, is not this announcement.
Check and explore
Sources for this article
Numbers connect each reference to the passages that use it. Dates show when the documentation was consulted.
This article draws on the sources above. The exercises are for you to try on your devices; SecuFocus does not present them as tests carried out by its editorial team. Interfaces and features can change. Method and corrections.
Cite this article
Keep this reference with the article when you save or share it.