API pricePriced at $0.08 per minute ($4.80 per hour) in the Voice Agent API; stated to be the same price as Grok Voice 2.0.
XAI · MODEL RELEASE TRACKER
Grok Voice 3.0
Grok Voice 3.0 is a speech-to-speech model from xAI, announced as available in the Voice Agent API. The announcement reports benchmark-leading scores on multiple audio benchmarks, sub-second latency, large-scale deployments (Tesla, SpaceX, Starlink), and API pricing of $0.08/min ($4.80/hr).CURRENT SNAPSHOT3/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
Not established from the available sources.
ModalityDescribed as a speech-to-speech model (marketed as "the world’s most intelligent speech-to-speech model").
Speech-to-Speech Index (Artificial Analysis)Reported #1 on Artificial Analysis's Speech-to-Speech Index with a score of 86.9 (Index combines speech reasoning, τ-Voice, and preference and task success from blind head-to-head calls).
Full Duplex BenchReported score of 97% on Full Duplex Bench, a measure of handling pauses, turn-taking, and interruptions.
τ-Voice scoreReported τ-Voice score of 71.6% (measuring handling of diverse accents, background noise, and interruptions); stated as comparable with GPT-Live-1 (Astra).
EVA-Bench accuracyReported highest accuracy on EVA-Bench (ServiceNow’s benchmark for enterprise voice agents) at 74% and described as the best-performing model when jointly scored across accuracy and customer experience.
Not established from the available sources.
VERIFIABLE FACTS
Every value stays attached to a source and date.
RELEASE · API availabilityDEVELOPER CLAIM
Announced as available today in the Voice Agent API.
PRICING · API priceDEVELOPER CLAIM
Priced at $0.08 per minute ($4.80 per hour) in the Voice Agent API; stated to be the same price as Grok Voice 2.0.
MODALITIES · ModalityDEVELOPER CLAIM
Described as a speech-to-speech model (marketed as "the world’s most intelligent speech-to-speech model").
CAPABILITIES · LatencyDEVELOPER CLAIM
Reported class-leading latency under one second; stated to have 40% lower latency than GPT-Live-1 (Astra) in comparisons.
BENCHMARKS · Speech-to-Speech Index (Artificial Analysis)DEVELOPER CLAIM
Reported #1 on Artificial Analysis's Speech-to-Speech Index with a score of 86.9 (Index combines speech reasoning, τ-Voice, and preference and task success from blind head-to-head calls).
BENCHMARKS · Full Duplex BenchDEVELOPER CLAIM
Reported score of 97% on Full Duplex Bench, a measure of handling pauses, turn-taking, and interruptions.
BENCHMARKS · τ-Voice scoreDEVELOPER CLAIM
Reported τ-Voice score of 71.6% (measuring handling of diverse accents, background noise, and interruptions); stated as comparable with GPT-Live-1 (Astra).
BENCHMARKS · EVA-Bench accuracyDEVELOPER CLAIM
Reported highest accuracy on EVA-Bench (ServiceNow’s benchmark for enterprise voice agents) at 74% and described as the best-performing model when jointly scored across accuracy and customer experience.
WHAT CHANGED