Transcription · xAI
xAI Speech to Text
xAI's speech recognition, the cheapest transcription in the catalog. Anonymized.
Anonymized Low costWhat sets it apart
xAI's speech to text is the cheapest in the catalog, at $0.10 per hour in batch. It covers 25+ languages, labels speakers, handles multichannel audio and formats numbers and entities. Version 2.0 roughly halves its error rate.
- $0.10 per hour in batch, per xAI
- 25+ languages, diarization, multichannel
- Formats numbers, currencies and emails
Recommended for: transcribe meetings and dictations.
Strengths and limits
Where it shines
- Lowest price per minute here
- 25+ languages, per xAI
Think twice when
- Anonymized, not private: xAI's retention policy applies
- Newer and less tested than Parakeet or Whisper
- 25+ languages
Use cases
Transcribe a phone call with names and amounts
Transcribe this customer call on two channels, format the amounts and email addresses, and label the agent and the customer.
It handles multichannel audio and entity formatting, per xAI.
Transcribe a large archive on a small budget
Transcribe these 40 recorded lectures as plain text, one file per lecture.
The per-hour price is the lowest here.
Against its rivals
- ScribeScribe covers 90+ languages and sound tags, at a higher price.#1 of 4 on Artificial Analysis word error rate (2.2%)
- WhisperWhisper is private, open and covers 99 languages.#3 of 4 on Artificial Analysis word error rate (4.1%)
- ParakeetParakeet is private and equally cheap, for 25 European languages.
Version history
Each version, dated, with what it changed over the one before.
Hollow points are earlier versions no longer offered in Sub Rosa.
See the numbers
| Version | Released | % |
|---|---|---|
| Grok Speech to Text (1.0) | 2026-04-17 | 4% |
| Grok Voice Transcribe 2.0 | 2026-09-18 | 2.3% |
- Grok Speech to Text (1.0)Latest
- Standalone speech to text API with batch and streaming, speaker labels, timestamps and multichannel audio
- Inverse text normalization for numbers, currencies and email addresses
- Grok Voice Transcribe 2.0Earlier, not in Sub Rosa
- Twice as accurate as 1.0 at the same price, per xAI
- Multilingual error rate on short phrases down from 20.6% to 6.8% across 19 languages
Benchmarks
How to read these scores: what a benchmark tells you.
Artificial Analysis word error rate: #2 of 4 models measured in the catalog. The share of words a transcription gets wrong (missed, added or replaced) on a mixed set of real recordings. Lower is better. At 5%, one word in twenty needs fixing; under 3% a transcript reads almost clean.
See the numbers
| Model | % |
|---|---|
| 1. ElevenLabs Scribe V2 | 2.2% |
| 2. xAI Speech to Text v1 | 4% |
| 3. Whisper Large V3 | 4.1% |
| 4. Wizper (Whisper v3) | 4.7% |
Announced by the maker
Measured by the maker with its own settings: useful, but not directly comparable with the independent scores above.
| Benchmark | Model | Score | Source |
|---|---|---|---|
| Artificial Analysis word error rate | xAI Speech to Text v1 | 6.9% | x.ai |
The model in this family
| Model | Use | Privacy | Price |
|---|---|---|---|
xAI Speech to Text v1stt-xai-v1 | Transcription | Anonymized | $0.048 per hour of audio |
Specifications
| Specification | Grok Speech to Text 1.0 |
|---|---|
| Released | April 17, 2026 |
| Weights | Closed |
| Accepts | audio |
| Produces | text |
| Languages | 25+ languages |
| Price | $0.048 per hour of audio |
| Privacy | Anonymized |
Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.
Sources
Everything on this page comes from these pages, read on October 16, 2026 or as dated on each score.