Grok Voice 3.0 Tops the Speech-to-Speech Index — Already Handling Real Calls at Starlink and Tesla

by Geth Author

xAI released Grok Voice 3.0 on October 14, 2026. The new speech-to-speech model takes the top position on the Speech-to-Speech Index with a score of 86.9 — according to xAI, citing Artificial Analysis — and ships at the same $0.08 per minute price as its predecessor. The grok-voice-latest alias now points to the new model automatically, so existing integrations pick it up without any code changes.

What changed under the hood

Grok Voice 3.0 is a true speech-to-speech model: audio in, audio out, with no intermediate text representation. This matters for latency. Independent measurements put time-to-first-audio at 0.81 seconds, compared to 1.34 seconds for OpenAI's GPT-Live-1. For voice agents handling customer calls, that difference is perceptible — 530 milliseconds of extra delay is enough to make a conversation feel sluggish. xAI also claims the model scored 74% on EVA-Bench, ServiceNow's benchmark for enterprise voice agent accuracy, though that number comes from xAI's own launch post and has not been independently verified at time of writing.

Already in production at Starlink and Tesla

The most concrete evidence of Grok Voice 3.0's capabilities is where it is already deployed. According to reporting from multiple outlets, the model is handling inbound calls at Starlink and Tesla — two companies within the xAI orbit — as of launch day. Voice agents in enterprise settings need to navigate interruptions, background noise, domain-specific vocabulary, and long multi-turn conversations without losing context. Real-world deployment at that scale is a more meaningful signal than benchmark scores alone.

Pricing that matches the competition

At $0.08 per minute, Grok Voice 3.0 is priced to compete directly with ElevenLabs and OpenAI's voice offerings. xAI has not raised the price from Grok Voice 2.0, which is an unusual move for a model claiming top-of-benchmark performance. The decision appears deliberate: API adoption is driven by predictable costs, and keeping pricing flat means teams already building on Grok Voice will not need to re-evaluate their unit economics before upgrading.

What this does not include

It is worth noting what Grok Voice 3.0 is not. This is an API-only release — xAI made no announcement about changes to the consumer Grok app or Grok's voice feature in X (formerly Twitter). The update is squarely targeted at developers building voice agents, call center tools, and interactive voice response systems. Consumer voice improvements, if any, will presumably come separately.

For teams building voice-first applications, Grok Voice 3.0 is now the benchmark leader on paper and has real production deployments to back it up. The unchanged pricing makes it an easy upgrade for existing Grok Voice users and a credible option for teams currently evaluating providers.