Universal-1

AssemblyAI has launched Universal-1, a multilingual speech-to-text model with exceptional accuracy and efficiency. Trained on over 12.5 million hours of audio data in English, Spanish, French, and German, it significantly improves word error rates and timestamp estimation, and supports code-switching.

• Over 12.5M hours of training data.
• 92.5% accuracy rate reached.
• Supports English, Spanish, French, German.
• Available via API for developers.
• Reduces hallucination rates by 30%.