SpaceXSI's Grok Voice 3.0 beats OpenAI, Google in speech benchmark
What's the story
xAI (now known as SpaceXSI) has unveiled its latest innovation, the Grok Voice 3.0. This state-of-the-art speech-to-speech model is designed to transform customer support interactions. It boasts an impressive record for resolving calls quickly and efficiently, with a latency of under one second. Notably, it also comes at a lower cost than other advanced models in the market.
Performance metrics
Model topped speech-to-speech index
Grok Voice 3.0 has been independently tested by Artificial Analysis, where it topped the Speech-to-Speech Index with a score of 86.9, beating OpenAI's GPT-Live-1 (83.2) and Google's Gemini 3.8 Live (82.6).
This index evaluates speech reasoning, t-Voice, preference, and task success from blind head-to-head calls.
The model also scored an impressive 97% on Full Duplex Bench, a test that measures how naturally it handles pauses, turn-taking, and interruptions in conversations.
Economic advantage
Model cheaper to run than other comparable models
Grok Voice 3.0 is cheaper to run than other comparable models, as per Artificial Analysis's Big Bench Audio run.
It comes in under both GPT-Live-1 (Astra, Medium) and GPT-Realtime-2.1 in terms of task completion costs.
The model has also shown its versatility by scoring the highest accuracy score of 74% on EVA-Bench, ServiceNow's benchmark for enterprise voice agents.
Practical use
Already in use in Tesla cars and Starlink call centers
Grok Voice 3.0 is already being used in Tesla vehicles for millions of conversations every day.
Drivers can use Connectors for simple requests like checking their calendar or answering an email, and Grok Bot for bigger tasks like ordering a coffee or booking dinner, all completely hands-free (see it on X).
It also powers Starlink call centers, answering 20,000 customer calls daily across multiple languages with an impressive resolution rate of around 70% without human intervention.