xAI announced Grok Voice 3.0 on October 14, 2026, making the speech-to-speech model available through its Voice Agent API at the same price as Grok Voice 2.0. The company says it resolves customer support calls at a record rate, has latency under one second, and costs less than other frontier models. Tesla and SpaceX use Grok Voice for millions of conversations each day.

In Artificial Analysis testing, Grok Voice 3.0 ranked #1 on the Speech-to-Speech Index with 86.9, ahead of Grok Voice 2.0 at 81.3, GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6. It scored 98.5% on Big Bench Audio for speech reasoning, 1,070 on Speech Agent Arena Elo for human preference, 94.8% task success, and 97.0% on Full Duplex Bench for conversational dynamics. The corresponding scores for Grok Voice 2.0, GPT-Live-1 and Gemini 3.8 Live were 97.2%, 90.1%, 97.7%; 1,011, 1,048, 990; 94.6%, 87.4%, 89.1%; and 95.1%, 94.9%, 91.9%. The company says Full Duplex Bench measures handling pauses, turn-taking and interruptions.

Grok Voice 3.0 scored 71.6% on τ-Voice, which tests support calls with varied accents, background noise and interruptions. xAI says that result is comparable to GPT-Live-1 (Astra) with 40% lower latency. Artificial Analysis found its cost per task on Big Bench Audio below GPT-Live-1 (Astra, Medium) and GPT-Realtime-2.1. On ServiceNow’s EVA-Bench for enterprise voice agents, it achieved the highest accuracy score at 74% and ranked best when accuracy and customer experience were scored together; the benchmark covers 200+ enterprise scenarios.

Tesla drivers can use Grok Voice for millions of daily conversations, including hands-free calendar and email requests through Connectors, and tasks such as ordering coffee or booking dinner through Grok Bot. Grok Voice 3.0 also powers Grok Bot’s voice mode. Starlink call centers use Grok Voice for 20,000 customer calls daily across multiple languages. About 70% are resolved without a human, and one in four sales calls ends in a purchase before the caller hangs up. A single agent can verify customers, troubleshoot hardware, issue credits and ship replacement equipment.

The Voice Agent API price is $0.08 per minute ($4.80 per hour), unchanged from Grok Voice 2.0. Agents using grok-voice-latest were switched automatically, with no changes required.

xAI announced Grok Voice 3.0 on October 14, 2026, making the speech-to-speech model available through its Voice Agent API at the same price as Grok Voice 2.0. The company says it resolves customer support calls at a record rate, has latency under one second, and costs less than other frontier models. Tesla and SpaceX use Grok Voice for millions of conversations each day.

In Artificial Analysis testing, Grok Voice 3.0 ranked #1 on the Speech-to-Speech Index with 86.9, ahead of Grok Voice 2.0 at 81.3, GPT-Live-1 at 83.2 and Gemini 3.8 Live at 82.6. It scored 98.5% on Big Bench Audio for speech reasoning, 1,070 on Speech Agent Arena Elo for human preference, 94.8% task success, and 97.0% on Full Duplex Bench for conversational dynamics. The corresponding scores for Grok Voice 2.0, GPT-Live-1 and Gemini 3.8 Live were 97.2%, 90.1%, 97.7%; 1,011, 1,048, 990; 94.6%, 87.4%, 89.1%; and 95.1%, 94.9%, 91.9%. The company says Full Duplex Bench measures handling pauses, turn-taking and interruptions.

Grok Voice 3.0 scored 71.6% on τ-Voice, which tests support calls with varied accents, background noise and interruptions. xAI says that result is comparable to GPT-Live-1 (Astra) with 40% lower latency. Artificial Analysis found its cost per task on Big Bench Audio below GPT-Live-1 (Astra, Medium) and GPT-Realtime-2.1. On ServiceNow’s EVA-Bench for enterprise voice agents, it achieved the highest accuracy score at 74% and ranked best when accuracy and customer experience were scored together; the benchmark covers 200+ enterprise scenarios.

Tesla drivers can use Grok Voice for millions of daily conversations, including hands-free calendar and email requests through Connectors, and tasks such as ordering coffee or booking dinner through Grok Bot. Grok Voice 3.0 also powers Grok Bot’s voice mode. Starlink call centers use Grok Voice for 20,000 customer calls daily across multiple languages. About 70% are resolved without a human, and one in four sales calls ends in a purchase before the caller hangs up. A single agent can verify customers, troubleshoot hardware, issue credits and ship replacement equipment.

The Voice Agent API price is $0.08 per minute ($4.80 per hour), unchanged from Grok Voice 2.0. Agents using grok-voice-latest were switched automatically, with no changes required.