On September 18 (local time), SpaceXAI, the artificial intelligence (AI) arm of SpaceX, officially rolled out its cutting-edge speech-to-text model, Grok Voice Transcribe 2.0. Leveraging the robust Grok Voice audio model, which is seamlessly woven into real-world business applications, this latest iteration significantly cuts down error rates by nearly 50% compared to its forerunner. Dominating the streaming model leaderboard on the third-party evaluation platform Artificial Analysis, it clinched the number one position for accuracy by a substantial margin, outperforming the 1.0 version across all four internal dataset evaluations. When it comes to pricing, Grok Voice Transcribe 2.0 sticks to its original pricing strategy, with batch transcription services costing $0.10 per hour and real-time streaming transcription priced at $0.20 per hour. Moreover, advanced functionalities like speaker diarization, precise timestamping, and custom keyword recognition are now part of the standard package, offered at no additional charge. The model’s commercial API is now accessible, and developers can delve into comprehensive technical documentation and integration guides via the SpaceXAI official website.