News
Deepgram Unveils Nova-3 Medical, Advanced Speech AI for Healthcare
- By John K. Waters
- 04/16/2025
Voice AI company Deepgram has unveiled Nova 3 Medical, a next-generation speech-to-text model built specifically for the healthcare sector, claiming it delivers the most accurate real-time medical transcription on the market.
Announced Tuesday at the HIMSS25 Global Health Conference in Las Vegas, Nova-3 Medical is designed to power advanced voice AI applications in clinical environments, where accurate and secure transcription of medical terms is critical to patient care and regulatory compliance.
The model outperformed competing products in benchmark tests, achieving a median word error rate (WER) of 3.45%—a 63.6% reduction compared to the next best solution. More notably for healthcare use cases, it recorded a keyword error rate (KER) of 6.79%, reducing critical term misrecognition by over 40%, Deepgram said.
"Nova 3 Medical represents a significant leap forward in our commitment to transforming clinical documentation through AI," said Scott Stephenson, CEO of Deepgram, in a statement. "We are empowering developers to build products that improve patient care and operational efficiency."
The model is built to address challenges posed by noisy or distant audio inputs and supports seamless integration with electronic health record (EHR) systems. With features like Keyterm Prompting for up to 100 customizable medical terms, on-premises and VPC deployment options, and full HIPAA compliance, the platform targets hospitals, telehealth providers, and healthcare technology developers.
At a starting price of $0.0077 per minute of streaming audio, Deepgram says the solution is more than twice as affordable as leading cloud-based competitors. The company's voice AI platform also includes text-to-speech (TTS) and speech-to-speech (STS) components, making it a full-stack offering for enterprise applications.
Nova-3 Medical is the latest addition to Deepgram's enterprise runtime platform, used by over 200,000 developers and responsible for transcribing over one trillion words across 50,000 years of audio. The company is positioning itself as a core infrastructure provider for a booming medical transcription market, projected to grow to $190 billion by 2032.
About the Author
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].