Pin Grok Voice Transcribe 1.0 on the Speech-to-Text API

pin-grok-voice-transcribe-1-0Updated 2026-09-22

Grok Voice Transcribe 2.0 is the default Speech-to-Text model. The Sep 18, 2026 news post says 1.0 will be deprecated in the coming weeks. To keep 1.0 during that window, pass the model id on every call.

News: Introducing Grok Voice Transcribe 2.0. Model table: Speech to Text.

REST

curl -X POST https://api.x.ai/v1/stt \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -F model=grok-voice-transcribe-1.0 \
  -F file=@audio.mp3

Put model (and any other options) before file.

Streaming WebSocket

wss://api.x.ai/v1/stt?model=grok-voice-transcribe-1.0&sample_rate=16000&encoding=pcm

If you omit model, the server uses grok-voice-transcribe-2.0.

When to pin

Pin when a regression test, customer SLA, or offline eval still targets 1.0 output. Plan a cutover: news says existing integrations that omit model get 2.0 accuracy with no code changes once 2.0 is the default, and pricing stays $0.10 / hr batch and $0.20 / hr streaming.

Pitfalls

A hardcoded omit-model path will move to 2.0 as soon as that default is live — pin if you must stay on 1.0.
Do not invent a third STT model id. Docs list only grok-voice-transcribe-2.0 and grok-voice-transcribe-1.0.
Pinning does not change price; both models share the Speech-to-Text rates on the pricing page.