Introducing Universal-3 Pro: A new class of speech language model optimized for Voice AI

Universal-3 Pro is a groundbreaking speech language model introduced by AssemblyAI, designed to optimize Voice AI applications by allowing adaptable transcription through user prompts. This innovative model enhances audio processing by enabling users to provide specific instructions for transcription, including capturing disfluencies, speaker labeling, and audio tagging, all without needing parameter configuration. The model is particularly notable for its ability to improve accuracy by 45% in domain-specific terms through keyterm prompting and supports six languages with built-in code-switching capabilities. Universal-3 Pro offers significant improvements over traditional methods by processing audio contextually, which ensures more accurate and reliable transcriptions out-of-the-box, and delivers these capabilities at a lower cost compared to competing solutions. The model is available for free trial throughout February 2026, offering an accessible opportunity for teams to evaluate its performance in real-world scenarios.