One-line summary
On October 15, 2026, SpaceXAI launched Grok Voice 3.0, a speech-to-speech model that the company says is the smartest of its kind globally, with sub-1-second latency and lower running cost than comparable frontier models; third-party evaluator Artificial Analysis ranked it first in its speech-to-speech index with a score of 86.9, and the API is priced at $0.08 per minute, unchanged from Grok Voice 2.0.
Event Overview
According to IT Home's report on October 15, 2026, SpaceXAI released Grok Voice 3.0 that day, which the company describes as the most intelligent speech-to-speech model in the world. SpaceXAI said the model set new records in customer-service call handling, has latency below 1 second, and lower running cost than other frontier models. Grok Voice supports millions of conversations daily across Tesla and SpaceX and is now available via API at the same pricing as the previous generation.
Key Information
| Item | Detail |
|---|---|
| Release date | October 15, 2026 |
| Publisher | SpaceXAI |
| Product | Grok Voice 3.0 (speech-to-speech model) |
| Independent benchmark | First in Artificial Analysis speech-to-speech index, score 86.9 (data of October 13, 2026) |
| Key metrics | Full-duplex benchmark 97%; speech reasoning (Big Bench Audio) 98.5%; task completion rate 94.8%; τ-Voice 71.6%, 40% lower latency than GPT-Live-1 |
| API pricing | $0.08 per minute ($4.8 per hour), same as Grok Voice 2.0 |
| Existing usage | Millions of in-car conversations daily at Tesla; Starlink customer center handles about 20,000 calls daily, roughly 70% resolved without human agents |