xAI Launches Grok Voice 3.0, Topping Speech-to-Speech Leaderboard
On October 15, xAI officially released the speech-to-speech model Grok Voice 3.0, calling it the world's smartest speech-to-speech model. Targeted at voice agent and customer support call scenarios, it is now available via API and the Voice Agent Builder, priced at $0.08/minute—the same as the previous generation and cheaper than other frontier models. It ranked first on Artificial Analysis's Speech-to-Speech Index with a score of 86.9, beating GPT-Live-1 (Astra, 83.2) and Gemini 3.8 Live (82.6).
Confirmed
- First audio response in under 1 second, with latency down 40% from the previous generation
- API pricing at $0.08/minute, now live on the API and Voice Agent Builder
- Score of 86.9 on Artificial Analysis's speech-to-speech index, ranked first
- Has autonomously resolved 70% of Starlink support calls internally at SpaceX
- @rohanpaulai noted the model has capabilities related to τ-Voice tool calling
Why it matters
- Voice agents are seen as a key deployment path for customer support automation; with sub-1-second responses and $0.08/minute pricing, Grok Voice 3.0 directly challenges OpenAI's and Google's real-time voice products
- Autonomously resolving 70% of support calls within SpaceX/Starlink provides real-world, large-scale deployment data, demonstrating the feasibility of voice models replacing human agents
2026-10-15 ~ 2026-10-15 · 15 related posts
Primary sources
- xAI launches Grok Voice 3.0 speech-to-speech model with sub-second latency — SpaceXSI · 2026-10-15
- [source] Grok Voice 3.0 cuts latency 40%, autonomously resolves 70% of Starlink calls — SpaceXSI · 2026-10-15
- [source] Grok Voice 3.0 tops Artificial Analysis voice index, priced at $0.08/min in API — SpaceXSI · 2026-10-15
- Grok Voice 3.0 Tops Artificial Analysis Speech Rankings at $0.08/Minute — XFreeze · 2026-10-15
- Grok Voice 3.0 tops Artificial Analysis speech index at 86.9, 40% faster than GPT-Live-1 — rohanpaul_ai · 2026-10-15
- Grok Voice 3.0 Hits 98.5% on Speech Reasoning Benchmark, Ranks #2 on Agentic Voice Arena — ArtificialAnlys · 2026-10-15
- Grok Voice 3.0 Priced at $4.80/Hour of Input Audio, 2.2x Cheaper Than GPT-Realtime-2.1 High — ArtificialAnlys · 2026-10-15
8 near-duplicate retellings: rohanpaul_ai · rohanpaul_ai · ArtificialAnlys · rohanpaul_ai · rohanpaul_ai · veggie_eric · minchoi · minchoi