SpaceXAI Launches Grok Voice 3.0 With Lower Latency Than OpenAI And Google Models
The pricing and latency claims matter because SpaceXAI says Grok Voice 3.0 costs the same as the model it replaces, so existing grok-voice-latest agents switch over without code changes.
Reporting from 1 source: GIGAZINE.
SpaceXAI announced Grok Voice 3.0 on October 14, 2026. In Artificial Analysis tests, each metric improved over Grok Voice 2.0 and surpassed GPT-Live-1 (Astra, medium) and Gemini 3.8 Live (Ext. Thinking, High). On the Speech to Speech Index it ranked first at 86.9. Input audio costs $4.80 per hour, the same as version 2.0, and the Voice Agent API costs $0.08 per minute.
Grok Voice 3.0 is already running inside Tesla vehicles and at Starlink call centers, and it reached the Voice Agent API on the same day it was announced. SpaceXAI says the model set a record in resolution time for customer support calls. On Big Bench Audio, the average time to first audio was 0.81 seconds, against 1.34 seconds for GPT-Live-1 (Astra, medium) and 1.35 seconds for Gemini 3.8 Live Extended Thinking (High). Cost per hour of input audio is unchanged from Grok Voice 2.0.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.