GamiWays
Cloud APICommercial

xAI Grok STT

STT haute précision — 6.9% WER global, 25+ langues, diarisation, ITN avancée, $0.10/h batch

200ms
Latency (best case)
400ms
Latency (typical)
6.9%
WER (general audio)
$0.0017/min
Price per minute

Comparative Scores

Accuracy (WER)8/10
Streaming latency7/10
Multilingual7/10
Sovereignty2/10
Price accessibility9/10
Streaming quality8/10

Architecture

ArchitecturePropriétaire end-to-end (xAI), même stack que Grok Voice Think Fast 2.0
ParametersN/A (cloud)
Languages25+
Self-hostable No
Streaming Yes
WER clean audio 2.3%
GamiWays
Phase B R&D — S2S pipeline evaluation

Medium relevance: competitive pricing ($0.10/h) and strong accuracy make it worth evaluating for Storygami/Edugami. However, US-only hosting and no GDPR documentation are blockers for Swiss/EU deployment. Best evaluated as part of the Grok Voice Think Fast 2.0 S2S pipeline ($0.08/min all-in) rather than standalone STT.

Analysis

Grok STT is the standalone transcription API built on the same stack as Grok Voice Think Fast 2.0. It achieves 6.9% WER globally across phone calls, meetings, video/podcasts and telephony — outperforming ElevenLabs Scribe v2 (9.0%), Deepgram Nova-3 (11.0%) and AssemblyAI (12.9%) in xAI's internal benchmark. Particularly strong on entity recognition (phone numbers, dates, currencies) thanks to advanced Inverse Text Normalization. Supports 25+ languages, word-level timestamps, speaker diarization, multichannel audio, and keyterm biasing. Priced at $0.10/h batch — among the most competitive in the market. Best evaluated as part of the Grok Voice Think Fast 2.0 S2S pipeline ($0.08/min all-in).

Strengths

  • 6.9% WER global — surpasse Deepgram, ElevenLabs, AssemblyAI (benchmark xAI)
  • ~10× plus précis en environnement bruité
  • ITN avancée — entités, dates, devises correctement formatées
  • Diarisation + timestamps mot par mot + multichannel
  • 25+ langues
  • $0.10/h batch — très compétitif
  • Même stack que Grok Voice Think Fast 2.0

Weaknesses

  • Benchmark interne xAI — validation Artificial Analysis indépendante à confirmer
  • Cloud US uniquement — pas de région EU, RGPD à vérifier
  • Pas d'option on-premise
  • Pay-as-you-go uniquement — pas de plan d'abonnement
  • Écosystème intégrations tiers moins mature que Deepgram

STT Capabilities

Streaming Yes

WebSocket streaming temps réel. REST API pour batch. VAD serveur intégré. Formats : PCM, PCMU, PCMA, Opus. Keyterms (100 termes max) pour biais de transcription.

Diarization Yes
Custom Vocabulary Yes
Word Timestamps Yes
Auto Punctuation Yes
Multilingual Yes

25+ languages

Pricing

Price / minute
from $0.0017
depending on plan
Price / hour
from $0.100
depending on plan
Free tier
Crédit de démarrage via console.x.ai (montant non publié)

Docs xAI du 17 octobre 2026 : REST $0.10/h, streaming $0.20/h. Modèle par défaut grok-voice-transcribe-2.0. Pas d’abonnement avec minutes incluses.

PlanSubscription/moIncludedOverage/min
StreamingRecommended

$0.20/h streaming. Batch file transcription is $0.10/h and is not this plan.

FreePay-as-you-go$0.0033

Sovereignty & Compliance

On-premise No

Cloud uniquement (US). Pas d'option on-premise documentée.

GDPR No

Data residency: US (xAI). Pas de région EU documentée. RGPD à vérifier.

On-premise No

Cloud uniquement. Pas d'option on-premise documentée.

This sheet's verification

Verified 17 October 2026

xAI internal benchmark (avril 2026) — validation Artificial Analysis indépendante à confirmer

Update note: AA-WER public 2,3 % pour Grok Voice Transcribe 2.0, relevé le 17 octobre 2026. Le 6,9 % interne d’avril et le tarif $0,10/h batch sont conservés.

This date covers provider-specific prices, capabilities and notes. Comparative benchmarks follow the synchronization and methodology shown above.

ELO benchmarks / indices: Artificial Analysis