xAI Unveils Voice AI 'Grok Voice Think Fast 2.0' with Dramatically Improved Transcription Accuracy and Inference Speed

U.S. artificial intelligence developer xAI announced its next-generation voice AI model, "Grok Voice Think Fast 2.0," on July 29. The model delivers substantial improvements over its predecessor in voice reasoning capabilities, conversational ability, and tool-use reliability, while also demonstrating superior performance as a transcription model. Compared to the previous model, "Grok Voice Think Fast 1.0," the new version significantly improves reasoning token processing efficiency. It achieves more advanced conversational capabilities while reducing reasoning tokens by approximately 60%, accelerating tool calls in production environments. In benchmark tests conducted by Artificial Analysis, Grok Voice Think Fast 2.0 demonstrated clear advantages over major competing models including OpenAI's "GPT Realtime 2.1 (High)" and Google's "Gemini 3.1 Flash (High)." The model recorded the highest scores or fastest results in four of five evaluation categories, with a time-to-first-audio speed of 0.70 seconds. The model also delivered standout results in transcription performance, achieving 1.5 to 2.0 times accuracy improvements compared to specialized speech recognition models across 24 languages. Migration to the new model will occur automatically for users specifying "grok-voice-latest" starting August 5, 2026, with pricing set at $0.08 per minute of audio.