AI Coding Live News

SpaceXAI (xAI) launched Grok Voice Transcribe 2.0, claiming roughly 2× the accuracy of Transcribe 1.0 on customer-support calls, spoken credentials, and short voice commands at the same price. It ranks #1 for accuracy among 32 streaming transcription models on Artificial Analysis; short-phrase WER fell from about 20.6% to 6.8%. Features include streaming and batch modes, automatic language detection with mid-recording language switches, word-level timestamps and confidence, free speaker diarization, batch at about $0.10/hour and realtime streaming at about $0.20/hour, with existing Speech-to-Text API clients upgraded without code changes.

Artificial Analysis (@ArtificialAnlys) updated its Speech to Speech Index: Google Gemini 3.8 Live Extended Thinking (High) debuts at #1 with 82.6, ahead of GPT-Live-1 (Astra medium) at 81.5 and Grok Voice Think Fast 2.0 High at 81.3; standard Gemini 3.8 Live lands #5 at 76.0. The same drop covers Speech Agent Arena, Tau Voice (Extended Thinking leads at 68.6%), Big Bench Audio, time-to-first-audio, and input-audio cost ($0.84/hr standard, $3.50/hr Extended Thinking).

Artificial Analysis (@ArtificialAnlys) published its Speech to Speech Index: OpenAI GPT-Live-1 with Astra (medium) as the delegated backend debuts at #1 with 81.5, while Sol (low) scores 80.1 at #3, with Grok Voice Think Fast 2.0 High between them at 81.3. GPT-Live-1 is a full-duplex speech model that can delegate reasoning and tool use to a backend text model mid-conversation. The same drop covers Speech Agent Arena, Tau Voice, Big Bench Audio, plus cost and time-to-first-audio.

Elon Musk quote-replied “New feature” to a user demo confirming SpaceXAI voice conversation mode can share your screen: the voice model sees what you do in the browser and talks you through tasks live. That upgrades Grok Voice from audio-only to visual coaching, useful for desktop coding help, debug walkthroughs, and longer agentic sessions.

SpaceXAI said Grok Voice models can now be used in LiveKit to build voice agents with full zero-data-retention support. LiveKit showed a patient-intake agent that cascades Grok STT, Grok 4.3, and Grok TTS through LiveKit Inference in a single AgentSession, with one set of credentials and billing. For developers, that is a three-string voice stack instead of a glued-together ASR, LLM, and TTS bill.

SpaceXAI said Grok Voice Think Fast 2.0 is #1 on the Artificial Analysis Speech-to-Speech Index, which scores whether voice agents can reason over speech, resolve real customer issues, and complete tasks with tools. The model is available via API and Agent Builder. SpaceXAI said Starlink is already using Grok Voice to handle more than 15,000 inbound support and sales calls a day.