xAI ships Grok Voice Transcribe 2.0 on speech-to-text API

xAI released Grok Voice Transcribe 2.0 on September 18, 2026, making the speech-to-text model available on its Speech-to-Text API for batch files and real-time streaming. Batch pricing remains $0.10 per hour of audio and streaming $0.20 per hour, matching version 1.0. xAI said existing integrations receive the change without code changes. The company called the model twice as accurate as version 1.0 and said it ranks first among 32 streaming models on the Artificial Analysis leaderboard. On an internal 19-language short-phrase set, it reported word error rate falling from 20.6% to 6.8%. Atlassian said it now transcribes every Loom video with the model. xAI said 2.0 will soon become the default and that 1.0 will be deprecated in the coming weeks.