SpaceXAI launches grok voice 3.0, claims it's the most intelligent voice model with sub-second latency

On October 15, 2026, SpaceXAI unveiled Grok Voice 3.0, positioning it as the most intelligent voice-to-voice model globally. The system achieved a sub-1-second delay and lower operational costs compared to other advanced models. It supports millions of conversations daily in Tesla and SpaceX vehicles, now available via API at the same price as its predecessor. Independent tests by Artificial Analysis ranked Grok Voice 3.0 first in a voice-to-voice benchmark with a score of 86.9, excelling in speech reasoning, τ-Voice metrics, and blind test preferences. It scored 97% in full-duplex benchmarks, measuring natural handling of pauses and interruptions. τ-Voice results showed a 71.6% score, matching GPT-Live-1 (Astra) but with 40% lower latency. Grok Voice 3.0 also outperformed GPT-Live-1 (Astra) and GPT-Realtime-2.1 in cost efficiency. In ServiceNow’s EVA benchmark, Grok Voice 3.0 achieved 74% accuracy, leading in combined precision and customer experience. It powers customer service in Starlink, handling 20,000 calls daily in multiple languages, with 70% resolved automatically. The model supports complex tasks like ordering coffee and booking dinner via voice, without manual input.