xAI details Grok Voice 3.0 benchmarks and customer-support use

xAI says Grok Voice 3.0 ranks first on the Speech-to-Speech Index with a score of 86.9.[1] The company says that index combines speech reasoning, τ-Voice, and preference and task-success results from blind head-to-head calls.[1] Grok Voice 3.0 scored 97% on Full Duplex Bench, which measures pauses, turn-taking, and interruptions.[1] A chart shows a 71.6% agent success rate, while a quoted result gives the model an accuracy score of 74% and says it sets a new Pareto frontier on EVA-Bench.[1]

xAI says Starlink call centers use Grok Voice to answer 20,000 customer calls a day across multiple languages.[1] The company says roughly 70% of calls handled with Grok Voice are resolved without a human in the loop.[1] xAI also says one in four Starlink sales calls ends in a purchase before the caller hangs up.[1] It describes one Grok Voice 3.0 agent as able to verify customers, troubleshoot hardware, issue credits, and ship replacement equipment.[1] These figures describe the company’s stated use and results; they do not establish that the same outcomes will apply to other call centers.

xAI says Grok Voice 3.0 costs the same as Grok Voice 2.0 and less than other frontier models.[1] A quoted figure lists $0.025 for Grok Voice 3.0, though the evidence does not specify what that figure is priced per.[1] xAI reports latency under one second and says the model is comparable with GPT-Live-1 (Astra) at 40% lower latency.[1] For teams assessing adoption, these claims make the price basis and the fit of their own workload important points to verify.