Grok Voice Transcribe 2.0 Becomes the Default With No Date. Pin 1.0 Now

Our verdict: if SpaceXAI's Speech-to-Text API runs in production for you, pin grok-voice-transcribe-1.0 today. SpaceXAI says Grok Voice Transcribe 2.0 "will soon be the default in the Speech-to-Text API" and that 1.0 "will be deprecated in the coming weeks". It gives no date for either. A silent default swap changes your transcripts even when the new model is better. Our directory rates 2.0 conditional: cheap and feature-rich, but every accuracy number is SpaceXAI's own. Our picks stay GPT-Transcribe for batch and GPT-Live-Transcribe for streaming, both recommended.

What happened

On 18 September 2026 SpaceXAI announced Grok Voice Transcribe 2.0.

• The default switch, verbatim: "Grok Voice Transcribe 2.0 will soon be the default in the Speech-to-Text API, and Grok Voice Transcribe 1.0 will be deprecated in the coming weeks. To stay on it during the transition, pin grok-voice-transcribe-1.0."
• No date: when we re-read the announcement on 18 October, it still gave no calendar date for the switch or the deprecation.
• Price unchanged: $0.10 per hour of audio for batch and $0.20 per hour for streaming, with diarization, timestamps and key terms included.
• Features: batch and streaming, word-level timestamps with confidence scores, speaker diarization, up to 8 channels, and up to 100 key terms per request. Numbers, dates, currencies, phone numbers and email addresses come back in written form, filler words are removed, and turn detection is built in.
• No 2.0 model id: the post names only the id to pin 1.0.

The accuracy claims are SpaceXAI's

SpaceXAI calls 2.0 "twice as accurate as Grok Voice Transcribe 1.0, at the same price". It says 2.0 ranks first for accuracy among 32 streaming models on the Artificial Analysis leaderboard, and that on its own short-phrase set (in-car commands and similar) word error rate drops from 20.6% to 6.8%. We have not checked that leaderboard entry or tested the model ourselves.

Why it matters

An unpinned integration will move to 2.0 on a day SpaceXAI has not named. Better average accuracy can still break things downstream: different speaker labels, different number formatting, removed filler words and new turn boundaries all change the text your parsers, search indexes and QA checks expect.

On paper the price gap against our picks is large:

Both OpenAI models stay recommended. The price gap is real and the diarization gap is likely, but we want independent accuracy numbers on audio like yours before we move a pick. Until then 2.0 stays conditional.

What changes for you

1. Pin 1.0 now in every Speech-to-Text API call that must produce stable output.
2. Test 2.0 on a sample of your own audio. Compare word error rate, diarization and key-term handling, and diff the formatted output against what your pipeline expects.
3. Look up the 2.0 model id on SpaceXAI's models page before you hard-code it. The announcement does not give one.
4. Plan the move before 1.0 is deprecated. Pinning only buys time until then, and that date is also unannounced.

FAQ

When does the default switch happen? SpaceXAI has not said. The announcement says "soon" for the switch and "in the coming weeks" for the 1.0 deprecation.

Does 2.0 cost more? No. Batch stays $0.10 per hour of audio and streaming $0.20 per hour.

Should I switch from GPT-Transcribe? Not on SpaceXAI's numbers alone. Run 2.0 on your own audio first; if it holds up, the price difference is worth acting on.